eCommerce

AI Agents Need a New Kind of Account Management in E-Commerce

It just dawned on me: account management in e-commerce platforms needs to be completely reimagined to support our AI agents.

Right now, account permissions are designed for people. The platform assumes you’re logging in yourself—or maybe that you’re sharing credentials with a spouse, a coworker, or a team. The permission model is binary: you either have the keys to the kingdom, or you don’t.

But what happens when it’s not you logging in, but an AI working on your behalf?

The First Step: Read-Only Access for AI Agents

Let’s start with the basics. I want to be able to give my AI agent read-only access to my account.

That could mean:

  • Checking my order status so I don’t have to dig through emails
  • Looking up past purchase history to remember what size I bought last time
  • Viewing personalized pricing or offers without being able to actually buy anything

This sounds simple, but it’s a big shift from today’s “all or nothing” credentials.

Beyond Shopping: AI as a Subscriber

For some sites, I don’t even need my AI to shop—I just want it to access what I’ve already paid for.

Take my America’s Test Kitchen membership. I love their human-driven product reviews—deep, professional, and credible. Right now, I log in myself and browse for recommendations. But imagine I’m shopping for a new food processor. I want my AI agent to:

  1. Access my membership
  2. Read the latest review roundup
  3. Recommend the best option for my needs

America’s Test Kitchen still gets my subscription revenue, but my agent does the work.

A Cascade of New Use Cases

This shift unlocks a flood of possibilities:

  • An AI travel assistant checking loyalty balances and upgrade offers
  • An AI home maintenance bot tracking past appliance purchases for warranty claims
  • A health and fitness AI that reviews my supplement orders to adjust my meal plan

Every one of these requires granular permissions—because I’m not ready to hand over the “buy now” button without limits.

The Double Bonus: Parental Controls

If we get this right, there’s another win: parental controls.
The same permission layers that protect me from rogue AI spending can protect kids from accidental (or intentional) purchases—while still letting them access the content or tools they need.

The Big Shift Ahead

We’ve been here before. When mobile apps arrived, we had to rethink authentication and permissions. When APIs took off, we created OAuth scopes. Now, AI agents are pushing us into another identity and access management evolution.

The companies that solve this first will have a competitive edge—not just in security, but in customer trust.

Because in the era of AI-powered commerce, “who’s logged in” won’t always be a human. And that changes everything.

Now, back to picking out a new food processor for my giant chimichurri recipes.

Business

AI for SMBs: The Real Bottleneck Isn’t the Tech—It’s the Data

If you’re trying to make sense of what AI can actually do for your business, you’re not alone. Every week, there’s another think piece, listicle, or research report telling entrepreneurs and business leaders how to leverage the latest wave of generative AI to scale. Most are written for the Fortune 500—not for teams still figuring out which SaaS tool will even talk to their QuickBooks.

This week, I came across a new Harvard Business Review article co-authored by one of my all-time favorite professors, Jeff Shay: How Ambitious Entrepreneurs Can Use AI to Scale Their Startups. Jeff was my professor back when I was earning my B.S. in Business Administration at the University of Montana, and we’ve stayed in touch ever since. He’s always had the right advice at the right time—like when he told me (long before it was cool) that “business plans are dead,” and pointed me toward the business model canvas instead.

His latest piece is aimed at founders and business leaders just starting to grasp the power of AI—laying out the core frameworks for where to look for automation and efficiency gains. It’s accessible, practical, and perfect for anyone new to the space. If you’re early in your AI journey, you’ll find it valuable.

But here’s where I’d build on it—especially for small and mid-sized businesses.

The real challenge most companies will face when implementing AI isn’t the technology itself. Tools and APIs keep getting easier. Multi-agent orchestration? Deep integration? These will be commoditized and packaged for the masses in no time—or the chosen AI platform will set itself up for you.

The real bottleneck is data.

If your data is scattered, unstructured, or just plain messy (think Excel files on someone’s hard drive), no amount of AI magic will save you. What most SMBs need isn’t just another AI tool—they need a new mindset around data. Data wrangling, data literacy, and an organizational culture that sees data as a business asset—these are the competitive advantages of the next decade.

I’ve seen it firsthand: from my early work at Adobe with Target and Recommendations, to wrestling Watson into commerce platforms, and now at Yottaa, the teams that win are the ones who invest in the “unsexy” stuff—cleaning, structuring, and understanding their data.

If you’re leading an SMB and thinking about how you’ll take advantage of AI, spend less time shopping for the flashiest new LLM and more time figuring out what questions you want answered and whether you have the data to get there. I’m a big fan of the Goal–Question–Metric approach: start with your business strategy, define the questions that reveal whether you’re making progress, and let the metrics fall out of that process. Those metrics will tell you what data you need—and whether it’s in a system an AI can actually use.

We’re close to seeing sophisticated prediction and attribution models made accessible to SMBs, but data science is still a field of trial and error. There’s a lot of potential math out there that could help a business leader predict outcomes or attribute effort to results—but it will require outside help. Over the next few years, that will likely come in the form of consultancies and contractors, until the discipline of data science becomes more common in SMBs. Given the typical IT adoption curve, it might take longer. In the meantime, get your Excel files organized and at least onto a shared drive.

Kudos to Jeff and the other authors for helping more entrepreneurs get on the field. The next level is building that data muscle. (And if you’re curious about Jeff’s take, check out his LinkedIn post here—)

Retail

Adaptive Content for Bots: Why Many E‑Commerce Sites Are Invisible to AI

Generative AI is rapidly changing how people search for and discover products online. As an e-commerce veteran, I’ve noticed a curious gap about to explode: our sleek, modern websites might look great to human shoppers, but to AI-based visitors they could be practically invisible. We optimize for Google and mobile devices, yet we now have to consider a new audience—AI crawlers and chatbots. It’s time to take a hard look at how our content appears to these bots and explore the idea of “adaptive content for bots” as the next evolution in digital strategy.

Current State: AI Crawler Limitations and Evidence

AI-driven crawlers (like OpenAI’s GPTBot or Anthropic’s ClaudeBot) are already hitting websites in huge numbers – one report noted GPTBot and ClaudeBot made roughly 939 million requests in a month, about 28% of Googlebot’s volume. In other words, these AI bots are a significant (and growing) part of web traffic. The catch? They’re often crawling with blinders on due to technical limitations. Consider the evidence and current limitations:

  • No JavaScript Rendering: Unlike Google, most AI crawlers do not execute JavaScript when they crawl a page. GPTBot, ChatGPT’s browser plugin, Claude’s bot – none of them run your scripts. They fetch the raw HTML and that’s it. If your site relies on client-side JS to load core content, the bot simply won’t see it. (OpenAI’s crawler will grab your .js files, but it won’t actually run them.)
  • Invisible Dynamic Content: Any content injected after initial page load – via AJAX calls, SPAs, React components, etc. – might as well not exist from an AI’s perspective. Many have tried this out: feed ChatGPT a React-based product page and you often get a blank or boilerplate answer. In one test it responded that “the content may be JavaScript-based and couldn’t be retrieved”. The AI essentially shrugged because it saw almost nothing. I tried to get the phone number on a business recently in giant font in the footer and Claude couldn’t see it.
  • Limited Interaction: These crawlers don’t click buttons, scroll, or wait for asynchronous data. They operate like a very primitive browser. That means features like infinite scroll product catalogs, content behind login or pop-ups, or user-triggered filters will not be navigated or understood by AI agents in their current form.
  • Blocking and Access Issues: On top of technical limits, many websites are outright blocking AI crawlers via robots.txt due to data usage concerns. According to PPC Land, over 35% of the top 1000 websites now block OpenAI’s GPTBot (up from just 5% a year prior). Major companies like Amazon quickly implemented such blocks when GPTBot first appeared. This trend reflects mistrust of AI scraping, but it has a side effect: if you block these bots, your content won’t show up in their models or answers. As one analysis put it, sites face a dilemma between protecting content and maintaining visibility in AI-driven search results.

In short, today’s AI crawlers have significant tunnel vision. They’re powerful, but they primarily read static HTML and often only what’s immediately available. Anything not in that initial payload might as well be invisible. That’s a stark contrast to Google’s crawler, which will actually render JavaScript in a second-wave indexing process. We’ve spent years building rich, interactive sites under the assumption that “Google will handle it” – but now along comes a new class of crawlers that don’t handle it.

Adaptive Content for Bots – The “M.Brand.com” Parallel

Historical Precedent: The mobile web revolution saw companies create separate “m.brand.com” sites optimized for mobile devices. Today’s challenge requires a similar solution: AI-optimized content delivery that creates automatically generated, bot-readable versions of JavaScript-heavy sites.

So how do we bridge this gap? This is where adaptive content for bots comes in. Just as we once embraced responsive design for different screen sizes, we may need to adopt a kind of “responsive content” strategy for different visitor types (human vs. AI). The goal isn’t to cheat or cloak content, but to ensure the AI sees the same key information that a human would – delivered in a bot-friendly way.

Think of it as creating a simplified, pre-rendered storefront for AI visitors. When an AI crawler (or a chatbot’s browsing tool) comes knocking, the site could detect that user agent and serve up content that doesn’t require a human browser to make sense. This might involve techniques like server-side rendering (SSR) or generating a static snapshot of your page content on the fly. In fact, SEO experts are already recommending server-side rendering of important content so AI bots can understand and index it. It’s the same idea as making a website accessible – but here our “accessibility” target is an algorithm rather than a screen reader.

A few guiding principles are emerging for adaptive content targeting bots:

  • Provide Full HTML Content: Ensure that all critical text (product names, descriptions, prices, reviews) is present in the initial HTML response. If that means rendering it on the server or at the edge, it’s worth it. The bot should not need to run any scripts to get the gist of your page.
  • Preserve Consistency: The content you serve to bots should match what users see. This isn’t about showing something different (which could be seen as deceptive); it’s about delivering the same information in a more digestible format. For example, a dropdown menu of specs could be expanded into a simple list in the bot-facing HTML.
  • Use Standard Signals: Leverage things like structured data (schema markup) and emerging standards (if any, like a hypothetical LLMs.txt file) to guide bots. While experimental, these can help ensure the AI understands the context of your content. At minimum, keep using schema for products, reviews, etc., since AI agents may pay attention to it much like Google does.

In practice, adaptive content for bots might mean your e-commerce platform, CDN, or middleware detects AI crawlers and serves them an unobfuscated version of the page. No fancy client-side tricks—just the goods, plain and simple. This way, when Perplexity or another AI reads your site, it actually finds what it needs to confidently include your products or information in its answers.

We have optimized for desktop, then mobile; we optimized for web accessibility; now we have to optimize for AI. It’s a natural next step in the evolution of content delivery. And given how fast AI-driven search is growing, it’s one we can’t afford to ignore.

Why It Matters for Online Retailers

You might be thinking: this sounds like a lot of technical fuss for bots—why should retailers prioritize this? The short answer is that AI-driven search and shopping are quickly becoming mainstream, and they have real revenue implications. Here’s why this trend should be on every online retailer’s radar:

  • Shifting Consumer Behavior: A significant chunk of consumers are already using generative AI platforms as a starting point for search and shopping queries. About 10% of U.S. consumers now prefer AI chat platforms for search, a figure projected to swell to over 90 million people by 2027. If even a fraction of those users are asking AI for product recommendations or answers, you want your brand to be in the response. Ignoring this channel could mean missing out on a growing audience.
  • Higher-Intent Traffic: Early data suggests that AI-referred traffic is especially valuable. Users who arrive via an AI assistant often spend more time and convert at higher rates than typical search visitors. One study by Zen Agency noted that AI-referred visitors stayed about 2.3 minutes longer and had ~1.5× higher conversion rates compared to regular search traffic. It makes sense – if someone asks an AI for “the best noise-cancelling headphones” and your product is recommended, that visitor is likely far down the purchase funnel and ready to act. Being invisible to these AI means losing out on some of the most qualified leads out there.
  • New AI Shopping Experiences: Major AI players are actively integrating shopping features. OpenAI recently announced a shopping mode in ChatGPT that can surface product picks with buy buttons embedded. Essentially, ChatGPT might become a concierge that points consumers directly to purchase links on retailer sites. But here’s the rub: if ChatGPT can’t properly read your site or doesn’t know about your products (perhaps because your content was hidden or you blocked the crawler), your items won’t be among those recommendations. Instead, it will favor products it can read about – possibly your competitors. Retailers need to ensure they’re not inadvertently closed off from these emerging AI-driven storefronts.
  • SEO Isn’t Enough Anymore: For years, we focused on climbing Google’s rankings. Now there’s a parallel challenge of getting noticed by AI algorithms (some call it Generative Engine Optimization or GEO). It’s not just about traditional SEO signals; it’s about feeding the AI the right data. If your beautifully designed site is essentially a blank page to an AI crawler, all the classic SEO optimizations won’t help in this new context. Being proactive about adaptive content ensures you don’t lose ground as search paradigms shift.

In essence, the rise of AI search means online retailers must double-check their digital storefront’s visibility. It’s no longer enough that a human with a web browser can navigate your site – now you have to consider the AI “visitor” who is blind to anything beyond raw text. The payoff for doing so is not just maintaining traffic, but tapping into a channel where recommendations carry extra weight and user engagement is high.

Looking Ahead

All of this paints a picture of a web in flux. We have a bit of a Wild West situation where AI models are reshaping how consumers find information, yet the infrastructure of the web hasn’t fully caught up. The good news is we’ve faced similar challenges before (remember the early mobile web, or the move to responsive design?) and we have the tools to adapt again.

Personally, I find this challenge exciting. It’s diagnostic and exploratory right now – we’re identifying the cracks in the system. As someone who’s spent a career at the intersection of commerce and technology, I know the solution is around the corner. I won’t delve into specific ideas here, but let’s just say there is a way to make adaptive content for bots dead simple for any e-commerce site to implement. My goal is to ensure retailers don’t need to overhaul their tech stack just to be seen by an AI.

For now, the key takeaway is awareness. Know that these limitations exist, audit your own site, and start thinking about how to serve AI crawlers the content they need. It might be as straightforward as flipping on server-side rendering for key pages, or as involved as using an edge worker to generate static snapshots. Whatever the approach, the first step is recognizing the gap.

The storefront of the future isn’t just what humans see – it’s what algorithms see as well. Adaptive content for bots could well become a standard practice in the years ahead. Or, this is irrelevant and the AI bots will read JavaScript by the time you read this old blog entry.

Business

From Digital Analytics to Business Outcomes Monitoring (BOM): Rethinking Composable Digital Experiences

A playful twist on “BOM,” traditionally known as a Bill of Materials, feels entirely appropriate given the explosive growth of digital experiences built from an ever-expanding set of vendors. Today’s e-commerce and digital business leaders aren’t managing single, monolithic systems—they’re orchestrating a vast array of technologies, each promising improved customer engagement and higher conversion rates.

But here’s the challenge: How do you measure performance, pinpoint issues, and optimize business outcomes in an ecosystem that often feels fragmented, overwhelming, and noisy?

The Journey Here: How We Got Composable

E-commerce software for selling online has evolved through three distinct phases:

  • Phase One: The Monolithic Era
    Companies chose “all-in-one” platforms like IBM WebSphere Commerce, Intershop, and later Magento, aiming for a single solution that delivered everything from product catalog management to checkout.
  • Phase Two: Headless Emergence
    The rise of the iPhone and social media changed the game. Suddenly, brands needed product information and commerce functionality not just on their website, but across mobile apps, social networks, and other digital touchpoints. Commerce platforms became “headless,” decoupling the storefront from backend operations. New specialized solutions emerged rapidly, commoditizing traditional e-commerce platforms.
  • Phase Three: Best-of-Breed Explosion
    Today, according to Gartner, “by 2026, at least 70% of organizations will be mandated to acquire composable digital experience platform technology, compared to 50% in 2023.” Companies large and small now actively mix vendor software, open-source solutions, and custom development to create highly personalized digital experiences. 

Composable digital experiences are no longer just a trend; they’ve become the new normal.

The Analytics Mess We’re In

As experiences got more sophisticated, analytics splintered. Companies now juggle at least three types of analytics solutions, each with its own perspective and agenda:

  • Digital Analytics: Dominated by tools like Google Analytics (GA4) and Adobe Analytics, focusing on consumer engagement and conversion metrics.
  • Performance Monitoring: Tools like Dynatrace, New Relic, and Catchpoint focus on technical IT metrics, such as server response times and app uptime.
  • Product Analytics: Companies like Amplitude and Quantum Metric focus on detailed user behaviors within digital products.

Each category generates massive amounts of telemetry data. But here’s the rub: How do these analytics translate into actionable business insights and measurable commercial outcomes?

Forrester introduced their “Digital Intelligence Stack,” recognizing the need to unify analytics data with actions that lead to improved business outcomes. Gartner labels this merging analytics field “Digital Experience Monitoring (DEM),” recommending that IT teams shift from purely technical metrics to business KPIs—such as customer satisfaction and quality of experience.

Yet even these frameworks leave a critical gap: translating metrics directly into action-oriented insights about vendor performance and measurable commercial outcomes.

How About Business Outcomes Monitoring (BOM)

Enter the ‘potential’ era of Business Outcomes Monitoring (BOM)—an integrated analytics approach explicitly designed for composable digital experiences. BOM focuses not just on data, but on connecting that data directly to business outcomes:

  • Vendor Performance Insights: Rather than just uptime or technical speed, BOM would pinpoint which vendor apps and integrations directly boost—or hinder—conversion rates and revenue.
  • Cross-Channel Optimization: Rather than siloed metrics, BOM continuously tracks and experiments across multiple digital touchpoints, allowing consistent, personalized experiences that lead to demonstrable commercial value.
  • Actionable Experimentation: Moving beyond A/B testing button colors, BOM could leverage sophisticated experimentation approaches (feature management and server-side testing) to validate larger, more impactful changes across customer journeys.

With composable architectures now standard and companies employing 50-100 vendor applications on their storefronts, real-time understanding of each vendor’s contribution to business success isn’t just helpful—it’s essential.

The Opportunity: From Data to Dollars

The shift from fragmented analytics to BOM creates a powerful opportunity. Gartner notes DEM can measure vendor SLA compliance, allowing businesses to demand service credits or prioritize investments based on direct business outcomes.

Early market signals suggest a ripe opportunity for vendors and analysts alike to rethink how we frame, measure, and manage digital experiences. Companies embracing BOM will inevitably build a competitive advantage—faster experimentation, precise investment, and clearer pathways to growth.

This is just the beginning of a significant shift—one I’m eager to explore further with peers, analysts, and industry experts alike.

How are you approaching this analytics evolution? I’d love your take. We’re working on this at Yottaa.

Technology

DeepSeek & The Future of AI Models: Lessons from the Browser Wars

The trajectory of innovation in AI today reminds me of the early days of the web and the browser wars that shaped it. Back then, our ability to create and interact with the web was constrained by a single browser, limiting the fidelity of what we could produce. As competitors entered the scene, standards like HTML and CSS emerged, and the web expanded exponentially. Even with those standards, new innovations like JavaScript and Macromedia’s Flash unlocked possibilities we hadn’t imagined, breaking through the constraints of what we thought browsers could do.

AI is heading down a similar path.

The Parallel to Browsers

We’re currently in an era where innovation in AI models is being driven by a mix of proprietary players (like OpenAI and Anthropic) and open-source challengers (like Meta’s LLaMA and the newly released DeepSeek R1). Much like the browser wars, this competition is defining early standards and pushing the boundaries of what’s possible. Proprietary models bring alignment, safety, and polished performance, while open-source models democratize access and drive experimentation.

But even as we’re watching this battle for dominance unfold, I think the focus on which model “wins” is irrelevant in the long run. Just as browsers settled into a multi-platform world with innovation layered on top, I believe we’re heading for a multi-model AI ecosystem, where different models—both proprietary and open-source—work together in a modular, interoperable way.

A Multi-Model World

The future isn’t about one model dominating. It’s about how we combine and integrate different models to meet specific needs. Proprietary models will likely dominate in commercial, consumer-facing products where safety and alignment are non-negotiable. Open-source models, on the other hand, will thrive in research, niche applications, and cost-sensitive projects, pushing forward innovation in areas that proprietary players might overlook.

What’s important here is that this trend line isn’t just about scaling existing architectures or improving fidelity. It’s about creating new tools, systems, and even paradigms—like JavaScript and Flash did for browsers—that will fundamentally shift how we think about AI applications. I’m convinced we haven’t yet seen the step changes that will truly unlock AI’s potential. These will likely come from new players or hybrid innovations that combine dense, sparse, symbolic, and neural approaches into something we haven’t imagined yet.

Standards, Middleware, and the Rise of Hybrid AI

If we take lessons from the browser wars, we can see a few trends emerging:

1. Interoperability Will Drive Adoption: Just as the web couldn’t scale without standards like HTML5, the AI space will need standardized APIs, formats, and protocols to allow models to work together seamlessly. Companies don’t want to be locked into a single model—they’ll demand orchestration layers that let them mix and match tools based on need.

2. AI Middleware Will Be Key: As multi-model workflows become the norm, we’ll see the rise of platforms that act as a gateway for managing and combining models. Tools like LangChain or AutoML are just the beginning—these systems will evolve to abstract away the complexity of choosing and managing models.

3. Hybrid Architectures Will Take Over: Just as browsers now integrate multiple engines (like JavaScript and WebAssembly), AI models will likely become hybrid systems that combine the strengths of dense networks, sparse MoE architectures, and symbolic reasoning. These modular systems will be able to scale while still delivering task-specific performance.

4. Innovation Will Outpace Standards (For Now): Before interoperability catches up, we’ll live in a fragmented space where innovation moves faster than standardization. This means we’ll see exciting breakthroughs but also challenges in integrating them effectively.

Irrelevance of Individual Models

Looking at this trajectory, it’s clear that the individual models themselves are becoming less relevant. What will matter more is the ecosystem: how models are integrated, how they interoperate, and how they solve real-world problems together.

The browser wars ended with multiple players coexisting because the innovation became more important than the competition. The same will happen with AI. Whether it’s DeepSeek, Meta, OpenAI, Anthropic, or the next unknown player, the trend line points to a modular AI world where tools work together seamlessly—and innovation, not dominance, drives the future.

The big question isn’t “which model will win?” but rather, “how do we build the tools and systems to make them work together?”

Business, Technology

SaaS companies are irrelevant with AI agents

I’ve heard a few folks joking (sometimes not-so-jokingly) about how the next wave of AI agents, combined with a simple database, could potentially wipe out the entire SaaS market. It’s a compelling vision: Imagine a single AI “super-tool” that knows your business inside and out, automates everything, and relegates your CRM, HR platform, and helpdesk software to the dustbin of history. Exciting, right?

But let’s talk reality for a moment. If you look at how technology has actually disrupted industries over the past couple of decades, it never really happens overnight. You don’t just flip a switch, dump all your existing systems, and pivot to a brand-new approach. Instead, it’s more like layering new capabilities on top of old ones, seeing what sticks, and slowly consolidating the tech stack over time. AI is absolutely going to transform the SaaS ecosystem, but “transform” doesn’t automatically mean “replace.” Let’s explore why.

As a technologist, I feel this coming. As a business person always surprised by the true pace of adoption, I know it’s a ways off.

Why SaaS Isn’t Going Away Anytime Soon—But How AI Agents Will Change the Game

The Slow Burn of Disruption

Adoption of cutting-edge tech usually starts at the edges, where the stakes aren’t as high. Think about when cloud-based software began replacing on-prem solutions. We didn’t see an immediate mass exodus. Companies tested the waters in non-critical areas first (like file sharing or team messaging), and only later migrated core systems. The same pattern will play out with AI agents.

Complexity and Compliance

On paper, “an AI + a database” sounds brilliant. But throw in industry-specific compliance, security protocols, and complex workflows, and you’ll realize how much engineering and domain expertise goes into even a “simple” HR or CRM platform.

The SaaS Value Prop Still Matters

Right now, top SaaS platforms offer more than just a place to store data. They’re infused with industry best practices and workflows that have been refined by thousands of customers over time. There’s a reason marketing automation tools or e-commerce solutions do specific tasks so well—the collective learning is baked in.

And then there’s compliance. If you run a business in healthcare, finance, or any highly regulated industry, your legal team needs robust evidence that a platform meets certification standards (HIPAA, SOC 2, GDPR, etc.). Trust me, “We used an AI that’s really cool” doesn’t cut it for an audit.

The Role of AI Agents

AI agents, particularly ones powered by large language models (LLMs), are already helping us automate tasks, generate insights, and connect data dots. However, in the near term, these agents will augment rather than completely replace your SaaS subscriptions.

The big challenge? Integration. Even the smartest AI can’t do much if it’s not talking to all your relevant data sources and apps in a secure, compliant way. And that integration layer—what I sometimes call “the plumbing”—is a huge part of SaaS value. AI might handle logic and recommendations, but the orchestration behind the scenes is massive. Even the MACH Alliance is finally starting to embrace the need for middleware.

Futures to Watch

  1. AI-Led Entrants: We’ll see new AI-powered platforms disrupt established players in specific niches—like specialized virtual assistants for customer support or data analytics that run circles around older solutions.
  2. Hybrid SaaS: Expect your favorite SaaS providers to fold AI features right into their core offerings. It’s already happening as they add GPT or other LLM-based functionality into their existing products.
  3. Specialized Micro-AI Services: Rather than one giant Skynet agent, we might see (and arguably already see) micro-AI components each tackling a narrow function—like invoice processing, lead scoring, or content generation—and snapping together like Lego blocks in your tech stack.

Timing Is Everything

When I first started evangelizing experience-driven commerce, which required headless commerce, I eagerly awaited all storefronts to change within a year or two. Back then I did not fully appreciating the cost of capital and the inertia to replace existing systems. With this in mind, I think the impact to SaaS will likely drag on for a decade.

  • Short-Term (1–3 Years): SaaS solutions get more AI smarts. Expect advanced natural language features, deeper automation, and chat-like interfaces on top of the same infrastructure you’re using now.
  • Mid-Term (3–7 Years): We start seeing the best-of-breed AI platforms truly rival (and sometimes replace) older, more rigid SaaS tools. If you’re running a simple or repetitive workflow, an AI agent might do it cheaper and faster.
  • Long-Term (7–10+ Years): By this point, many simplistic SaaS solutions may well be superseded by “AI + data” systems. But for mission-critical or highly regulated operations, specialized SaaS platforms—likely with strong AI capabilities of their own—will still be essential.

Final Word

As much as I love the sci-fi idea of a single AI brain running your entire business, we’re not quite there yet—and we won’t be for a while. The SaaS model still brings real value in terms of domain expertise, compliance, and that invisible but crucial layer of data plumbing. And, the reality is most business users just want to get to their kids softball practice and will drag their adoption feet.

That said, the future is definitely AI-powered. Over the next decade, expect a continuous blurring of lines between what we call “SaaS” and what we call “AI platforms.” In the end, though, the old question remains: who can deliver a product that solves real business problems, aligns with security and compliance needs, and fits into evolving workflows? AI is going to shake things up, but SaaS isn’t going away—it’ll just look a little (or a lot) different.

We’ll all keep an eye on those AI developments, that are just getting started.

Marketing

Personalization @ Scale Includes Privacy as a Priority

Recently I had the honor of being a guest speaker in the Customer Analytics course, which is part of the University of Virginia’s Master of Science in business analytics (MSBA) program. The class is taught by Professor Ryan Wright, a longtime friend and hero of mine in academia (he never ceases to amaze me on how well he has the pulse of industry and is preparing his students to thrive in it).

One of the students asked my position on privacy within the context of personalization and whether or not I thought we have the right regulations in place. I joked a bit that if most people knew what I knew, they’d flush their phone down the toilet and disconnect everything. But in reality, it comes down to an exchange, a barter if you well. A company gives you free services (e.g., email or this platform to post my thoughts) or a more relevant service (e.g., search results or pre-filtered category pages on an e-commerce site) in exchange for insight and information about you. Now, if this exchange comes with a degree of trust and transparency or better yet, control, then you are more inclined to exchange some data about yourself for the service. I am a big proponent of this equation, as I believe many of us want to continue to take advantage of free services and better experiences. However, I do believe that while things are rapidly expanding in regard to data regulations, we have a long way to go before it’s a fair exchange.

My strategy for personalization is to lead with privacy in the forefront, while acknowledging people want you to make their lives easier. Do I want the kid’s sections of a mobile app or website to pre-filter to my children’s sizes, tastes and closet contents? Absolutely! That of course means I have to be ok with that same company keeping track of my kid’s sizes, tastes and closet contents. Now here’s where the key is in the exchange. If the business provides the right level of transparency AND control, then I am comfortable. Lucky for me, I continue to find my way to companies that share my sentiment. Beyond the standard privacy policy, Gap Inc. has gone so far as to adopt privacy principles to guide internal projects. We’re also currently recruiting for an internal privacy counsel (plug to join Gap, Inc).

To help employees understand key Privacy issues and risks, Gap Inc. has adopted Seven Privacy Principles to help guide projects and initiatives. 

1.) Consent: Obtain permission from individuals before using their personal information.

2.) Control: Give individuals meaningful choices about how their information is used.

3.) Fairness: Use personal information in ways appropriate for the context.

4.) Minimization: Collect and store personal information only as needed to provide the service.

5.) Confidentiality: Securely store and transfer personal information and only share it when necessary.

6.) Access: Allow individuals to correct or delete their personal information.

7.). Accountability: Socialize and enforce these principles.

So how do I intend to bring this to life for our brands? Going back to the “exchange”, in my opinion the barter works best when the company creating the service based on the data gives the customer ridiculously easy to find (transparency) and operate controls.

Google

Let’s step into a couple examples to illustrate the concept and show that some of the ‘big bad wolves’ actually do a decent job of this and have created a precedent in my opinion. Google for example now has most of what they do with your data consolidated into one place, with simple language as to what it is, how it’s used and really easy toggles to turn things on and off.

No alt text provided for this image

Nike

If you have a Nike account, you’ll notice they’re very up front about how they want to make the experience more personalized for you and allow you to help provide input to the process. This is a good balance between letting the modern new marketing tools figure out what you’re thinking with the old school process of simply asking, “Darin, what’s your shoe size?”. If you have a Nike account, you can see how they let you opt in or out to the personalization. When setup an account via their mobile app, they do a great job of explaining things in plain English and letting you decide.

No alt text provided for this image
No alt text provided for this image
No alt text provided for this image

Macy’s

While most of the customer preference center best practices have been about managing email communications, I also really like what Macy’s has done with their customer preference center. They allow you to proactively tell them your favorite sizes, brands, etc. Ironically though, after I did update mine and returned to the home page, it didn’t seem like it took my instructions into account.

No alt text provided for this image

In Closing

With great power comes great responsibility. As digital marketers, we need to thoughtfully balance the power of all our new machine learning algorithms and massive ability to collect enormous amounts of data on our customers with the tools and experiences we build to empower our customers to let us know what they want us to use.

eCommerce

My Headless Commerce Experience with Adobe.com

I’ve worked in commerce for some two decades, on both the front-end and the back-end and everything in between. Over time, I’ve honed a vision for what commerce should be. I believe many of you share a similar view. Yet we struggle to deliver given the limitations of existing technology patterns and vendor solutions. In the same way I’ve challenged myself throughout my career, I challenge the industry to materialize this vision. Here’s the story of my work on Adobe.com

/var/folders/94/z7nrsy593hn5sqwnwhr3xm094lwyq3/T/com.microsoft.Word/WebArchiveCopyPasteTempFiles/0?e=1557360000&v=beta&t=t7HoHBhcXGerKSCyNwdUFVhtE36_g3g2e2ukjag2lzw

A New Vision for Commerce

As shoppers, most of us are into one product category or another and some of us are into more than our finances support. For me I love gear. I love electronics gear, outdoor gear, sailing gear, hunting gear. Anything that acts as an aid to my hobbies, whether truly required or valuable, I’m all over it. Some of you may be into hand bags or collectibles, shoes or toys. Whatever your shopping vice, we can probably agree that the research and shopping portion of the journey are just as wonderful as the purchase itself.

Which is why I have always been disappointed with online shopping.

If I’m trying to buy a raincoator fancy backpacker style tent online, I want more information on why there is such a broad selection of products. What are the differences? Why is one item better than another? Retailers have been just plain lazy in how they merchandise online. Their typical, lack-luster responses to my questions include “faceted search” and ”sorts” on price or reviews.  Even manufacturers struggle to articulate why they created so many similar products in a category. We all put a lot of love and passion into product development. It’s just seemed crazy to me that we’re still mostly showcasing them as thumbnails in rows and columns online.

While at Adobe we encountered the typical challenge where we created these beautiful marketing pages and then customers had to go to an (ugly) ecommerce site to buy once they hit that point of inflection. Our goal back then was to radically shift our business from the channel to direct sales. It was obvious after looking at the analytics and going through the buy flow, we couldn’t have two sites. We had to transform that beautiful marketing site into the store.

We needed experience-driven commerce.

Who Gets Experience-driven Commerce Right?

The technology pattern that allows marketers to create any compelling experience they want and then make them shoppable has been a CMS, which in turn calls a headless commerce solution via an API. It seems simple enough, yet most of the ecommerce vendors out there are running on really old architecture patterns where the system is designed to deliver a home page, category pages, product detail pages, cart, and checkout…aka, the web shop.

When we attempted to transform Adobe.com with ATG (now Oracle Commerce) and Day Software’s CQ5 (now Adobe Experience Manager), we quickly found out ATG’s API wasn’t complete enough and had too many holes to bother with. Unfortunately, we purchased ATG as we thought we might acquire the company later. They were bought by Oracle and continue to pursue a web shop model, although they have improved their API (ten years later).

During this large program, called Adobe.com Transformation, we also had evaluated Elastic Path.

Everyone was enamored with Elastic Path. The developers loved it and marketing did too, as we knew we’d be able to weave together our dream experience. I actually think our idea of combining “content and commerce” with the CMS and ecommerce stack was largely influenced by Elastic Path’s capabilities and a key architect that joined us from Amazon (he put me on to microservices before the term was coined). We also had a lot of fun settling debates in meetings quoting Linda Bustos from getelastic.com There’s a great case study by Harvard Business Review that outlines this journey we took at Adobe, so I won’t expand on it more.

Once we implemented our vision for Adobe.com, making the site the store, we saw immediate results. Our conversion rates went up and it was not only noticeable in the analytics, but also in the actual sales reports. We were taking the same traffic and making a lot more money… In a few years Adobe.com went from a little over USD $200M annually to nearly $750M. The site now generates over USD$1B (Internet Retailer Top 100).

I began evangelizing the idea of experience-driven commerceand searching for the technology pattern required to pull it off. I switched business units and joined what became the Digital Marketing BU. My charter was product strategy and marketing for the idea, with the added goal of finding a commerce engine to acquire that would allow Adobe to deliver this capability.

While we started out dancing with Hybris (which was later sold to SAP… a story for another time), we eventually evaluated and collaborated with most of the ecommerce platforms on the market. All but one struggled to deliver full capabilities as an API so we could manage the shopping experience in Adobe Experience Managerand take advantage of other products such as Adobe Target, Audience Manager, etc.

Here’s what we found:

ATG’s API was incomplete with version 10 and only as of the latest version (11.3) is there a full REST API. But, it’s effectively on an EOL path with a novel attempt at a re-write, and Oracle’s new vision misses the point about customization. Oracle Commerce Cloud, the cloud version of ATG, limits customizations to webhooks and still focuses on the web shop pattern.

Hybris’ API was designed mostly for a mobile app. You could get product information out of it and engage the cart through checkout, but we had a lot of customers that ran into problem after problem trying to make it work. Simple use cases like being able to request a product from a staging environment were not supported by the API. The API only knew about products in the live production site. Hybris is now owned by an ERP company that created the reason for ecommerce to be a separate solution in the first place.

Magento had an old SOAP interface that had a poorly performing Rest API duct-taped on top. Even with the 2.x changes it’s still better suited for small volumes and simple web shops.

Digital River didn’t have an API yet.

Demandware wasn’t sure what to make of “experience-driven commerce”. They liked their model of forcing customers into web shop templates. Ironic given their target audience was fashion retailers with beautiful products. (I always wondered what Kate Spade really thought of her web store). This SaaS version of Intershop, now called Salesforce Commerce Cloud, was initially appealing to business owners, but they soon run into all sorts of brand problems as they try to choreograph experiences across digital touchpoints.

Intershop posed similar challenges to Demandware and we later learned that it essentially was Demandware (I’m pretty sure Salesforce didn’t know this when they bought it, but hey, they can just keep buying solutions until they get one that works for their customer base.)

IBM WebSphere Commerce came close with a recent REST Level 1 API, but it was similar to Magento in that it was a “RESTified” SOAP interface and required a great deal of expertise to use. So much so that the partner helping to integrate Adobe and IBM gave up on a couple of areas and tried to just make the argument that WCS should drive those pages (and thus the side by side pattern emerged – even though this breaks the first guiding principle of enabling marketing with a single tool for experience).

And then there was Elastic Path. Unapologetically API first. Not a REST-ified SOAP interface like WCS and Magento. Not a partial API that misses checkout like Salesforce. Not a mobile app-only API like Hybris that misses simple use cases like being able to test multiple versions and environments. Elastic Path had engineered the most advanced, and thoughtful API for marketers and commerce professionals to bring to life their ideas of enabling any commerce transaction at that point of inflection where they have excited customers enough to buy—ether by pumping the tongue of a shoe, calling out from the shower or walking into a store with a wearable device.

Elastic Path had already skated to where the puck was going (they’re Canadian, so I figured they were trained on this concept more than others). They had already envisioned these use cases and were excited to show how easily they could enable any client application or shopping experience whether through a CMS or custom app.

And the thing that stuck with me over the years? Elastic Path just worked. It worked for really large transaction companies like Symantec who wanted to embed commerce in their own applications and it worked for smaller companies across any industry. Unfortunately, sometimes you don’t get to buy or implement what the team really wants due to the normal “complexities” of a large company, but I was hooked on the idea of headless commerce.

Results

 

What about the Internet of Things and ordering through Alexa?

The smartphone gave rise to IoT. Just like the dot.com boom inspired one of history’s largest telecommunications build-outs, which later provided the bandwidth for the web to deliver the experiences marketers originally envisioned in the 90’s. Smartphone adoption paved the way for inexpensive compute inside small packages with wireless capability. Thanks to Apple, Google and the Android operating system, the bill of materials (BOM) for a device to have a sophisticated OS, great compute power and wireless connectivity dropped to a level where entrepreneurs’ could invent pool sensors that can tell you when to add chemicals.

Recruited away from Adobe by Intel to define retail solutions that would take advantage of this upcoming re-imagining of everything around us, I took a detour deep into the historic Silicon Valley. During which I learned a hell of a lot more about silicon and the integrated circuit than I would have ever imagined.

It was while working at Intel on IoT products and collaborating with many innovative retailers that I realized we not only needed to have experience-driven commerce in our mobile apps and websites, but that everything around us was going to become electronic commerce.

I worked with retailers trying to actually figure out how to identify a customer in the store and just let them walk out without going through the point-of-sale (self-checkout). We collaborated with toy manufacturers that wanted to have the box or toy itself somehow interact with a digital deviceto buy accessories (same use cases in B2B for replenishing devices in life sciences companies). There were companies with vending machines (Best Buy in airports) and many others creating new, and strange, pizza ordering experiences.

Obviously the convergence of physical and digital experiencesbecomes a reality over time (though costs still must come down). While my assertions had expanded and I had a focus on bringing the digital into the physical realm, one company kept delivering on this commerce vision, Elastic Path.

Today my aim is to help other digital commerce practitioners accomplish similar results and create e-commerce experiences we’ll all love.

IoT, Marketing

Moving “Digital” to the Physical Realm

As marketers, the pressure in recent years has centered on shifting to an all-digital approach and delivering a compelling customer experience across the vast number of digital touchpoints. Just as we’re leaving more traditional marketing channels behind, the advent of the Internet of Things (IoT), connected/ smart devices, and mobile technologies is blurring the lines between the digital, traditional and physical at a rapid rate. And with the excitement of nearly anything – in either the digital or physical realm – becoming a potential marketing vehicle, there also comes the possibility of a resurgence in data silos and fragmented customer experiences. Join Darin Archer, head of ‎IoT Industry Solutions Retail/CPG at Intel Corporation, as he explores the possibility of breaking down the barriers between disparate channels, data sets and back-office systems – and the transformational, connected marketing experiences made possible as a result.

 

 

Marketing

Ramblings for University level Principles of Marketing Class

I’d go for the disruptive approach. Marketing has so fundamentally changed. It’s all about digital marketing now. When Nike learns that it can have the same “reach” with their own digital online event as a Superbowl ad, you know we’re on a new path. Today’s it’s about engaging fans, building an audience and content marketing. It’s no longer people sitting on madison ave dictating a brand campaign for next year and taking half a year to put it together. Red Bull spends more on sponsorships, events and publishing than traditional ads (wait, a drink company has a leading magazine?). Rolex makes more money on events than watches.

Traditional Advertising

TV is dead

More young folks watch people play video games online than CBS. Oh, and these same video game players can fill a stadium easier than most bands today. But, it’s also evolving. It’s going to be more about the content and the device and knowning who’s watching than creating generic ads for massive audiences. Don’t believe me? Your setup box knows what you’re surfing on your iPad while sitting on the can and can change the ad you see when you get back to the TV (http://www.adobe.com/products/auditude.html). DVR’s were just the beginning. Now people binge watch Breaking Bad on Netflix and catch up on new shows via Hulu. This means that video ads will become more about speed and personalization. We’ll need more creatives and copy editors to come up with more versions tailored to very specific demographics.

Outdoor billboards

Notice how many are switching to digital displays. This is the first step to these all just being converted to the Google Ad network. It won’t be about a local sales force or agency buying locations to have a canvas hung for a month rather these screens will know who’s driving by them and who will be driving by them shortly. They’ll then be optimized to show ads to reach the most lucrative potential audience. This will be done by software algorithms and ad bidding networks. Might as well lump this into the SEM (search engine marketing) bucket.

Radio

Who listens to radio anymore? Spotify, Soundcloud, Pandora, Rdio and of course Apple dominate our cars and our earbuds. Honda just announced support for Apple’s CarPlay and Android Auto. If I were ClearChannel, I’d try to figure out how to buy Spotify and partner up with Google. Again, it won’t be about long drawn out campaign planning.

Print

The bookstore closed and most magazines are struggling to keep their numbers up. I can have a greater reach and real conversion with display, social or search ads. Now, it’s not to say these won’t be there, especially in our digital magazines on our Apple Newstand, but this will all merge with the display networks from Google, Apple, Bing, etc.

If I’m focused on B2B, I get more out of content marketing (webinars, emails, blogs, social) than any traditional media. I also use these new strategies and my own digital properties to build demand and measure the results. It no longer makes sense to spend money on something I can’t measure.

If I’m B2C focused, I’m all about getting close to my customer, building fans and exciting my base. Again, this is about content marketing, not Madison Avenue.

The agencies are all struggling and consolidating to hold on to revenue. Most have build digital businesses as they know they won’t be making tons of money on campaigns anymore rather they’ll be producing content. Even the ones that have done well picking up the task of online ad buying and optimization see that this will be done by software (Adobe Media Optimizer, Marin Software) and not humans.

So what do you focus on?

Analytics – understand how to read data, find audiences, determine trends and target. Psychographics dominates and demographics is worthless. Tomorrow’s marketers also have to be able to show where they see an opportunity using data and then measure their results and constantly optimize their efforts

Testing – the art of a/b testing is now multivariate. Understanding statistics and what “statistically significant” means is relevant to knowning if their idea is good or dead.

Creative – The fun stuff still exists and more marketers should get more comfortable piecing together campaigns themselves because they’ll have tools where they can take digital assets and put together something (with a good copy editor) that can be pushed out to their audience in real time (from their phone while on the way to work)

I’d have them all take $50 and build an online campaign for a trinket they sell on etsy or an ebook they publish on Amazon. If they have a friend that makes something cool or has a small business, even better. Get in and see how you can target audiences on Facebook, understand what a Pinterest buy button might mean to marketing (yep, they’ll be measured by sales tomorrow not just reach and impressions).

 

 

 

Verified by MonsterInsights