To choose an AI consulting firm, an ecommerce brand should name one business problem first, shortlist firms that have worked with the same stack (Shopify, Klaviyo, Meta Ads, Google Ads), ask each for a system it has running in production, and start with a short paid phase that has a written stop point. Price matters less than whether the firm checks the data before it builds. This guide is from Brand Media Works, an AI consulting and marketing data engineering firm for ecommerce brands.
What kinds of firm sell AI consulting?
Six kinds of firm sell AI consulting to a brand of $5M to $200M, and each is strongest at a different part of the job. The table sets out what each usually does well and what to check before hiring one.
| Type of firm | Usually strongest at | What to check |
|---|---|---|
| AI consultancy or boutique | Scoped builds led by senior people | Whether it has connected ecommerce data before, and who will do the work day to day |
| AI or software development firm, often with offshore or nearshore teams | Engineering capacity for custom builds | Whether discovery tests the business case, or goes straight to a build |
| Data engineering consultancy | Joining Shopify, ad and email data into one warehouse | Whether it can take the work past reporting into AI tools, or will hand over at the warehouse |
| Marketing agency with an AI practice | Channel and campaign knowledge | Whether the AI work is scoped and priced separately from media or retainer work |
| Large strategy or accounting consultancy | Programs across many functions and countries | Fit for a mid-market budget: one AI development firm describes a large consultancy's AI scoping engagement as 8 to 12 weeks, priced at $250,000 and up, with associates running the day-to-day |
| Fractional AI lead | Setting direction and choosing vendors without a full-time hire | Who builds what they recommend: one AI software and strategy firm puts a fractional Chief AI Officer at $2,000 to $8,000 a month |
The labels overlap, and many firms call themselves all of these. What separates them for an ecommerce brand is whether they have worked with the systems the brand runs on, and whether their first step is the data or the build.
What should an ecommerce brand check first?
- Ecommerce work, not an ecommerce page. In the Brand Media Works competitor register, 16 of the 74 AI consultancies and AI development firms had at least one page written for ecommerce or retail brands when we read them between 12 and 21 October 2026, even though the register was built partly from ecommerce searches. An industry page is not a project. Ask for one that used a brand's orders, ad spend and email data, and for a client you can speak to.
- Data before models. Gartner predicts that through 2026, organizations will abandon 60% of AI projects unsupported by AI-ready data, and found that 63% of organizations either do not have or are unsure whether they have the right data management practices for AI. A firm that proposes an agent before checking whether Shopify, Meta and Klaviyo report the same revenue is skipping the step that decides most outcomes. Data engineering for ecommerce covers what that step involves.
- Systems in production, not demos. IDC research reported by CIO found that 88% of observed AI proofs of concept did not reach wide-scale deployment: for every 33 a company launched, only four did. Ask what the firm has running today, how long it has run, who owns it and what it changed. Client results on firm websites are rarely audited, and claims such as a proof-of-concept success rate several times the market average seldom name a source.
- The people on the project. Ask for the names and roles of the people who will do the work. One development firm staffs each pilot with four to five people: a project manager, a solution lead, AI engineers and specialists such as a data engineer.
- The running cost. Ask for a monthly running cost at the brand's own volumes alongside the build price. What an AI agent costs to build and run shows why the two are different numbers.
What should you ask on the first call?
The first call should tell you how a firm thinks, not only what it has built. The table sets out seven questions and the difference between a strong and a weak answer.
| Question | Strong answer | Weak answer |
|---|---|---|
| What would you do first, and what would you not build yet? | One bounded problem, a check of the data behind it, and a list of things to leave until later | A platform or several agents before seeing the data |
| Which ecommerce systems have you connected? | Named projects with Shopify, Klaviyo or the ad platforms, and a reference | An industry page and logos with no project detail |
| What is running in production today? | A system, how long it has run, who owns it and a measured change | Demos, prototypes and pilots only |
| How will we know it worked? | A business metric and a test set with a pass mark, agreed before the build | Accuracy figures with no test set behind them |
| What will it cost to run each month? | A running-cost estimate at the brand's volumes | A build price only |
| What do we own at the end? | Code, prompts, test sets and data, in the brand's own accounts | Access to the firm's platform for as long as the contract runs |
| Where can either of us stop? | A written stop point early in the build | One contract for everything |
If the model will see customer records, add the questions in how to choose a private LLM partner, which covers where the model runs and who can see the data.
How should the first engagement be structured?
Buy a short, paid first phase, then one bounded build. Published assessments take one week at one automation consultancy, two to three weeks at one development firm and two to four weeks at one AI software and strategy firm, which says a standalone assessment is commonly priced at $7,000 to $35,000. The assessment should end with a written plan the brand could act on with any firm.
Ask whether the assessment fee comes off the build. One development firm credits 50% of the fee toward the setup fee of a build that starts within 30 days, and one automation consultancy deducts its assessment from the build price. Then keep the first build small:
- Weeks, not months. One automation consultancy says a first build should be working in two to six weeks, and that a first build planned to take six months has the wrong scope. One development firm runs pilots to a fixed budget over 30 days.
- A stop point in the contract. One AI consulting boutique writes a week-three stop point into its contracts: if it concludes the system should not go to production, it says so and refunds the remainder.
- A pass mark agreed up front, tied to the business metric the brand named on the first call.
How much AI implementation costs sets out published prices for each stage, and how we work describes the audit, build and handover steps Brand Media Works follows.
Which answers should make a brand walk away?
- A solution before a question. The firm proposes a platform, a model or an agent before asking how the brand's data is set up.
- No ecommerce project it can describe in detail, only an industry page.
- Results with no baseline, such as a percentage improvement with no before figure, method or owner.
- No names for the people who will do the work.
- No running-cost estimate before the build is signed.
- The firm keeps the code, prompts or test sets, so leaving means starting again.
What does an AI consulting firm cost?
Most firms do not publish prices. AI consultant cost in 2026 sets out what the firms that do publish charge, and how much an AI consultant costs for a $20M ecommerce brand sets those prices against a typical AI budget. Custom AI tools covers the kind of build Brand Media Works delivers after the first phase.
What does an AI implementation consultant do?
What is an AI implementation consultant explains how the role differs from an AI strategy consultant: one delivers a working system in production, the other a roadmap. It compares the timelines and prices firms publish for each, and sets out which one an ecommerce brand usually needs.
Should a mid-market brand hire a boutique or a large consultancy?
Boutique AI consultancy or a large consultancy compares the two using the large firms' own published government rate cards, which put a six-week team at well over £100,000 before anything is built. It sets out what a $5M to $200M brand gets from each, when each is the right choice, and how to compare quotes like for like.
What does an AI consultant actually do?
What does an AI consultant do? sets out the work from assessment to handover, the four common types of AI consultant and published rates. It also covers when an ecommerce brand needs one and when it does not.
Frequently asked questions
Should an ecommerce brand hire a large consultancy or a boutique?
For a brand of $5M to $200M with one or two problems to solve, a smaller firm that has worked with the same ecommerce stack is usually the better fit. Large consultancies suit programs across many functions and countries, and their scoping alone can cost more than a mid-market brand's first build.
How many firms should we talk to?
Enough to compare answers to the same questions. Send each firm the same one-page brief naming the problem, the systems involved and the metric, ask the seven questions above, and compare the first-phase proposals side by side.
Do we need an AI strategy before hiring a firm?
No. A named problem and access to the data are enough to start. A good first phase produces the plan, ranks the options and says which ones not to build yet.