E-commerce is the SMB category where the AI landscape has fragmented most aggressively. Every Shopify or BigCommerce store I look at has 6 to 12 AI-enabled apps installed, half of which are barely used. The app store is full of products promising "AI-powered everything" and most owners can't tell which ones are real and which ones are wrappers around a single OpenAI call.
This post is the working frame for AI in e-commerce SMBs. Where the app store is good enough, where custom agents win, and the four workflows that pay off most often.
I cover the general custom vs off-the-shelf frame separately. E-commerce is one of the categories where the hybrid pattern is most pronounced and most worth doing carefully.
Two things make e-commerce AI different from generic SMB AI.
The platforms (Shopify, BigCommerce, WooCommerce) have rich APIs and good app ecosystems. Many AI workflows are well-served by an app. The build vs buy default leans further toward buy than in most SMB categories.
The volume and seasonality are extreme. A store can do 50 orders one Tuesday and 5,000 the following Tuesday during a peak. The AI workflows that absorb peak load without requiring peak staffing are disproportionately valuable.
The combination means most e-commerce SMBs should default to using the platform's apps for commodity workflows and selectively building custom for the specific places the apps don't fit.
Honest list of e-commerce AI workflows where the app store solution is usually fine.
Generic chat support for top-of-funnel questions (shipping, returns, sizing). Tools like Tidio, Gorgias, and Re:amaze with AI features handle this well.
Product description writing for standard catalogs. Copy.ai, Jasper, and the platform-native AI writers produce acceptable descriptions for most products.
Basic upsell and cross-sell recommendations. The platforms (especially Shopify) have built-in recommendation engines that are good enough for most SMBs.
Email subject line and body generation for routine campaigns. Klaviyo, Mailchimp, and Omnisend have AI-assisted drafting that works fine for promotional emails.
Generic SEO copy and meta tag generation. Plenty of tools do this. The cost-benefit of custom is hard to justify.
If your e-commerce AI work is concentrated in these categories, do the boring thing and use the apps. Custom builds here would cost 10x and produce slightly worse results.
The workflows where the app store leaves real money on the table.
The seasonal pattern is what makes this special. During Q4 or your category's peak, your support inbox triples. The team can't keep up. Customer experience suffers exactly when revenue is highest.
A custom triage and drafting agent that handles the 60 to 70% of tickets that are routine (order status, returns, exchanges, basic product questions) at peak load while leaving the harder cases to humans is a clean win for e-commerce.
The custom version beats the app version when your support voice matters, when you carry specific product knowledge, or when your return and exchange policies have nuance the generic app can't reliably handle.
Build cost: $25,000 to $40,000. Payback: usually one peak season for a store doing seven figures in Q4 revenue.
I cover the general AI agents for customer support playbook separately. The e-commerce angle adds the seasonality calculation.
For stores with hundreds or thousands of SKUs, keeping product data clean is a real ops problem. Categories, attributes, search tags, filter facets, all need to be consistent and accurate.
A custom agent that reads new product data (vendor feeds, manual entries, supplier sheets) and applies your specific taxonomy is a high-impact workflow. The app store solutions are too generic; they don't know your category hierarchy.
Build cost: $20,000 to $35,000. The win is invisible to customers but compounds in search relevance, filter accuracy, and merchandising agility.
The agent reads sales velocity, returns data, and basic external signals (weather, season, related-category trends) and surfaces SKUs that need reorder action, markdowns, or merchandising attention.
Not a pure AI workflow; it's analytics with a language layer. But the language layer matters because the output goes to a busy operator who needs the recommendation phrased clearly, not a chart they have to interpret.
Build cost: $25,000 to $45,000. Payback comes from fewer stockouts and fewer markdowns, both hard to attribute precisely but real.
A custom agent reads support tickets, identifies questions that recur because the product page is unclear, and drafts updates to the product description, FAQ, or sizing chart that would have prevented the question.
Indirect but high-impact. The wins compound month over month because the deflection persists. Stores that run this workflow well see steady declines in support volume per order over time.
Build cost: $15,000 to $25,000. Hardest to attribute payback but often the highest-impact in the category.
I cover the product feedback loop frame elsewhere; this workflow is the e-commerce-specific implementation of that loop.
A mature SMB e-commerce AI stack typically looks like this.
Platform-native AI features: turned on for the commodity workflows. Description writing, basic recommendations, generic chat for top-of-funnel.
A specialized SaaS for support workflow: Gorgias, Re:amaze, Zendesk depending on the team's preference. AI features used as the default for tier-1 support.
A custom triage and drafting agent for peak-season support load. Optional during off-peak, business-critical during peak.
A custom product taxonomy and categorization agent. Always running, especially during catalog refreshes.
Optional: a custom inventory and demand agent if the SKU count and complexity justify it.
Optional: a deflection content agent for stores at high enough volume to justify the work.
The portfolio mixes platform AI, SaaS AI, and custom AI based on the workflow. Nobody is "all custom" or "all SaaS" in this category and ships well at SMB budgets.
Inventory the AI you already have. Most stores have AI in 4 to 8 apps they're already paying for. Turn on the features that look promising. Measure adoption for a month. Drop the apps where the AI features aren't used.
Spend a peak season with the platform AI on and notice where it falls short. The failure modes during peak are where the custom case lives. Don't build custom before you've felt the pain.
Pick the highest-pain custom workflow for your store and scope it tightly. The peak-season support triage is the most common answer. Inventory or taxonomy is the next most common.
Budget for build and operating in the right range. A $25,000 to $40,000 first build, $200 to $500 a month operating, and 4 to 8 hours a week of staff time in the first quarter. If a vendor is quoting you wildly outside these ranges, get a second opinion.
Build one agent that works in production before considering a second one. The portfolio approach is the destination, not the starting point. The starting point is one working agent.
The e-commerce category has the most product market fit for custom AI of any SMB category I work with, partly because the platforms are open enough to make integration cheap and partly because the operations questions are sharp enough that AI has real things to do. Done well, an e-commerce SMB at the right scale can absorb two to three custom agents and run them comfortably with no incremental headcount. Done badly, the same store will accumulate 12 apps and a vague sense that AI is "not really working". Pick which version you want to build.
RELATED READING
When SMB owners ask me about AI agents, the first thing I usually have to do is unscramble the word "agent". The term is used to mean five different things in five different vendor pitches, and most…
Most small business owners I talk to are stuck in the same place with AI. They've watched ChatGPT do something impressive. They've signed up for two or three SaaS products with "AI" in the name.…
Customer support is the most over-pitched and most under-shipped category in SMB AI. Every vendor wants to sell you a chatbot that "handles" 80% of tickets. Most of the chatbots SMBs deploy this way…
Back-office operations is the most underrated category for SMB AI investment. The vendor pitches focus on customer-facing flashy stuff. The real ROI sits in the unglamorous workflows: invoices,…
The question "where to start with AI in your business" is asked badly more often than it's asked well. I've heard it phrased as "what AI tool should we buy", "should we hire an AI person", "what's…
An efficient and easy product feedback loop is an essential component of every successful product. Your target audience should know how to reach out to you.
FREQUENTLY ASKED