Aranda & Co. / Who we work with

A part-time head of AI for e-commerce and DTC brands

This page is for the owner of an e-commerce or DTC brand with a small team, a storefront on Shopify or something like it, a support inbox that never empties, and a marketing spend that has to be defended every month.

The pitch you're getting right now is replacement: a service that does your agency's job or your support team's job for a fraction of what you pay today, usually sold on a resolution percentage or a return on ad spend with no named customer behind it. The newer version names a worry about AI search and sells an audit to settle it. We spend two to three weeks inside your company working out where your team's week goes between tickets, listings and campaigns, and what each of those costs you. You get a ranked list and a 90-day plan you own. If you want us to, we run that plan with you.

Book 20 minutes How it runs here

What owners ask us for

Support replies drafted, a person sends

To start, we'd use AI to tag incoming tickets and draft a reply in your tone from the order record and your policy. A person reads each draft and hits send. Most e-commerce support is a short list of repeat questions: where's my order, how do I return this, is this in stock. We'd track time per ticket and first-response time, and count how many drafts go out unedited.

A returns bot that stays inside your policy

Owners want the customer-facing bot everyone is selling, and it's the last thing we'd turn on. Version one answers order status and returns questions only, and it reads from one policy document that a person maintains. Any other question goes to a human. We'd count how many clicks it takes to reach a person. Each week a person samples its answers and counts the wrong ones. Customers who come back with the same question count against it too.

Product copy at catalog scale

The number we'd try to shrink is the time from new product to live listing. Writing titles, descriptions and attributes for every SKU is slow work that nobody loves, so copy gets thin or goes stale. The smallest useful version drafts each listing from a spec sheet and your brand guide, and a named person approves it before it goes live. The scorecard logs errors caught in review and return reasons that trace back to a description.

One email flow rebuilt with tools you already pay for

Your email platform probably already has AI features you're paying for and not using. We'd rebuild one flow with them, usually abandoned cart or post-purchase. The features draft the copy, and a person sets the segments and checks both before anything sends. The test is revenue per recipient on that flow against the old version. Unsubscribes matter too, and so does how long the rebuild took.

Ad creative drafts your media buyer tests

We'd have AI generate copy and concept variations from the ads that already work. Your buyer or agency picks what gets spend. Agencies sell AI creative as a way to cut production cost, so cost per creative produced is worth tracking. So is the share of AI-assisted creatives still running after the test window.

Being found when a shopper asks an AI

AI search and agentic commerce are real, and they're also the newest thing to be scared into buying an audit over. The first step is unglamorous. It's clean, complete, structured product data in your feeds and on your pages, which your marketplace listings need anyway. Progress is measured in feed errors and attribute completeness by SKU. The last check is one you can run yourself: ask an AI assistant about your own category and see whether your products show up.

How it runs here

AI assessment

In an e-commerce company the assessment means sitting with the support lead and whoever answers tickets, the person who runs email and ads (or the agency that does), the person who handles fulfillment and returns, and you. We follow the flows that eat the hours: a ticket from arrival to close, an order from placement to delivery, a return from request to refund, a new product from spec sheet to live listing, and a campaign from brief to send. The data we ask for is the helpdesk export with tags and handle times, order and return records, your platform's analytics, the invoices for every agency and app subscription, the support macros and the written policies the team follows. Out of that comes your top five uses for AI ranked by payback, with the hours and the dollars behind each, and a 90-day plan with a shortlist of tools and builders.

Weeks 1 to 3

The 90-day plan

What usually ends up on a 90-day plan for a brand this size is the work behind the storefront, where a mistake is cheap: ticket tagging and drafted replies, a rewrite of the returns and shipping policy so any future bot has one source to read from, catalog data cleanup, and one email flow rebuilt with the features you already pay for. What usually waits is the customer-facing bot, AI pricing, anything that touches checkout, and replacing an agency, because each of those needs the earlier work in place and a way to catch errors before a customer sees them. The plan says who owns each item and when it's due, and it's yours whether or not we keep working together.

Yours to keep

Fractional Chief AI Officer

Week to week the retainer is the scorecard: each item on the plan, who owns it, when it's due, and whether it's on track, in one page every week. The vendor work in e-commerce is mostly sorting out whether the AI add-on in your helpdesk or email platform does the job before you buy a standalone tool, reading app contracts for usage billing and who owns the data your bot is trained on, and writing a scope for any outside builder with a date attached. The policy work is a plain-English AI-use rule for your team: what the bot may and may not say, that a named person approves copy and that no review is ever generated, how a customer is told they're talking to AI, and a kill switch we have tested. We train the support team and the marketing person on the tools, and once a month we sit with you for a leadership review.

Monthly

What you get

Your top five uses for AI, ranked by payback
A 90-day plan with a shortlist of tools and builders
Written scopes for any outside builder, with dates
A one-page weekly scorecard with owner, due date and status
A plain-English AI-use policy for your team
A monthly leadership review

Tools we'd look at first: ChatGPT, Gmail, WhatsApp, Meta Ads, Google Ads.

Fixed fee for the assessment, agreed before we start. Monthly retainer after, if it fits.

Where it goes wrong

The bot states a policy you do not have, and you are bound by it

Air Canada's chatbot told a customer he could buy a full-price ticket and apply within 90 days for a bereavement refund, contradicting the airline's actual policy. Air Canada argued the chatbot was a separate legal entity responsible for its own actions; the tribunal rejected that, found negligent misrepresentation and awarded damages plus costs. That is the standard a returns bot now works under.

Source: Moffatt v. Air Canada (BC Civil Resolution Tribunal, 2024 BCCRT 149), summary

AI-written listings go live with nobody reading them

In January 2024 Futurism reported Amazon listings titled "I'm sorry but I cannot fulfill this request it goes against OpenAI use policy" and a description containing the placeholder "[task 1], [task 2], and [task 3]." Futurism read them as sellers generating copy without proofreading, and Amazon said it had removed the listings.

Source: Amazon Is Selling Products With AI-Generated Names Like "I Cannot Fulfill This Request It Goes Against OpenAI Use Policy" (Futurism)

AI-generated reviews now carry federal civil penalties

The FTC's final rule on fake reviews, announced August 14, 2024, bans reviews that misrepresent they are by someone who does not exist, such as AI-generated fake reviews, or by someone with no actual experience of the business, and lets the Commission seek civil penalties against knowing violators.

Source: Federal Trade Commission Announces Final Rule Banning Fake Reviews and Testimonials (FTC)

Questions to ask any vendor, including us

When the bot answers a returns or refund question, where does the answer come from, and what stops it stating a policy we do not have?
What is the measured resolution rate for a named client in our category, and how many clicks does it take to reach a person?
Does AI-written copy or a review ever go live without a named person approving it, and how do you prove every review is from a real buyer?

Questions owners ask

how much does a fractional chief ai officer cost for an e-commerce brand

The assessment is a fixed fee agreed in writing before any work starts, so you know the number before you say yes. The retainer is monthly and starts after the assessment, if you want us to run the plan. Outside builders come in on a written scope you've seen, and if an assessment doesn't fit your company we'll say so on the first call.

can ai replace my customer support team

Some of the work, yes; the team, we wouldn't plan on it. Repeat questions with a clear answer in the order record can be drafted or handled by a bot, and a person is still needed for anything with judgment, a refund decision, or an angry customer, and for catching the bot's mistakes. We plan support work so the path to a human is short and measured, and we'd rather you keep a smaller team you trust than rehire in a hurry.

should i use ai to write my product descriptions

For first drafts, often. The two conditions we'd put on it are that the draft starts from a real spec sheet so it can't invent a feature, and that a named person reads every listing before it goes live. Copy that nobody read is how placeholders and apology text end up on a public product page.

do i need an ai chatbot on my shopify store

Not on day one. A customer-facing bot can bind you to what it says, so it needs a policy document to read from, a tested handoff to a person, and someone sampling its answers every week. We'd put it later in the plan, after the behind-the-scenes work has shown the team how the tools behave.

we already pay for ai features in our helpdesk and email platform, do we still need help

Maybe the help you need is turning them on. Part of what an assessment checks is which paid-for features nobody set up, and the first items on a plan are often configuration and training with no new purchase. Where you do need something new, you'll get a written scope and a shortlist, and you can take that to any builder.

what is agentic commerce and do i need to get ready for it

It's the idea that a shopper's AI assistant will find, compare and eventually buy products on their behalf. Nobody knows yet how much buying will move that way or when, and the preparation that holds up either way is clean, complete, structured product data, which you need for marketplaces and search already. We'd start there and skip the audit.

Start with twenty minutes.

If your support inbox or your product catalog is where the hours go, book twenty minutes and tell us about it. Erick will ask where the time goes and what you've already tried, and you'll leave the call with the one thing he'd look at first.

Book 20 minutes