Kwame MensahIn conversation
In conversationLondon and Accra

AI your shopcan actuallyuse.

Kwame Mensah is a fractional head of AI for small online shops. From London and Accra he helps brands with ten to fifty people ship AI search, support and forecasting that their teams keep using. We asked him how. The last question in this interview is yours.

Kwame Mensah standing at a packing table in a small online shop's studio, daylight from a tall window
Plate 1
01The job

Two days a week, inside the shop.

What does a fractional head of AI do all week?

I sit inside a small online shop two days a week as its head of AI. With the founder I decide which jobs a model should take on and which it should never touch, then I ship them with the team. I work with what the shop already runs, usually Shopify, and I never resell a tool.

When a shop is ready, I help it hire its own head of AI and step back. That is the goal from the first week.

How did you end up doing this?

I studied computer science at KNUST in Kumasi and operational research at the LSE. Then five years as a data scientist at Kaya Market, a marketplace in Accra, and four as head of data at Northlight Home, a homeware shop in London. Since 2023 I have worked fractionally, for one or two shops at a time.

Kwame Mensah at a long table with a shop's small team, laptops open between product boxes
Plate 2
02The work

Three shops, three numbers.

Show me something you shipped.

Every project starts with one number we agree on, measured before I start. Here are three.

Lumi Skin AI search in Hackney, London

Customers typed things like "something for dry skin that won't sting" and got nothing back. We replaced keyword search with search that understands what people mean, trained on the shop's own product notes.

14% to 3%searches that found nothing, within one quarter; revenue from search up 22%

Shelves of skincare bottles in a small London studio, a staff member packing an order
Plate 3

Adwoa Prints support on WhatsApp in Makola, Accra

Most orders arrive on WhatsApp. We built an assistant that answers order and delivery questions in English, says plainly that it is an assistant, and hands over to a person in one tap.

61%of questions answered without a person; median first reply from 5 hours to 4 minutes

Hale Coffee forecasting in Bristol

A weekly forecast for every bean, so the roaster buys green coffee for the weeks ahead.

40%fewer stock outs in six months

A woman photographs printed fabric with her phone in a busy textile shop in Accra
Plate 4
03The price

A free call, then three ways to work.

What does it cost?

Everything starts with a free 30 minute call. After that there are three ways to work, and Accra clients can pay in cedis.

First call
30 minutes, to see if AI can move your number at all
Free
Audit
Two weeks on your search, support and stock data; a written plan ranked by money and risk
£4,800
Fractional
Two days a week as your head of AI, three months at least, then month to month
£6,500 a month
Build sprint
Six weeks to ship one thing, search, support or forecasting, with your team
from £18,000

Who is it for?

Brands with ten to fifty people, usually selling between one and fifteen million pounds a year online. If the audit finds that AI will not move your number, I say so and you keep the report.

04The limits

Where AI should never answer for you.

Where should a shop never use AI?

Refunds, anything about allergies or medicine, and any message that pretends to be a person. Every assistant I ship says it is an assistant and hands over to a person in one tap.

Customer data stays in the shop's own accounts. Assistants never see card details, and each project gets a one page note on what data it uses and why.

Sacks of green coffee in a Bristol roastery, a laptop open on a workbench
Plate 5
05Your questions

The last question is yours.

Write the next question on the empty line. A script typesets each answer from Kwame's notes on this page; for a real answer, book the call.

Editor's notes