Published . 3 min read.

How do you get senior leaders to actually use an internal AI tool?

By , Group AI Marketing Executive at Archetype Group, Ho Chi Minh City

Short answer

Build it around work people already resent, so it beats habit. Design for the reluctant user, with the prompt carried by the system rather than the person. Ground it in your own material. Then onboard leaders before everyone else, because a tool introduced by a regional lead survives and one introduced by a group email does not. This is how I approached ABI at Archetype Group.

Generative AI reached Archetype Group the way it reached most firms: as twenty separate experiments. Offices pasted client material into consumer chatbots, prompts lived in private notes, and nothing one country learned reached the next. The brief I was given was not "adopt AI". Every firm says that, and most mean a pilot that quietly ends. The brief was to make a tool survive contact with people who have deadlines. That is a product problem and a behaviour problem at once, and the behaviour problem is harder. The case study is at ABI.

Status, stated plainly

ABI is pre-pilot. As of September 2026 the plan is a pilot in October 2026, a beta in November, a business-development workshop in December and a launch in January 2027. The only results so far come from an internal demo and from the enablement work described below. Nothing in this post claims the tool is live or adopted across the firm, and the demo figures are not production numbers.

1. Find where the time actually goes

Before building anything I mapped the repeated work: proposal boilerplate, translating technical scope into client language, and reformatting the same content for twenty markets. A new tool has to compete with habit, and habit is strong. It wins by attacking tasks people already resent, because the incentive to change already exists.

2. Build for the reluctant user

ABI was designed around the person who does not want to learn a new tool. There is no prompt engineering at the point of use. The user picks a structured entry point for one of the jobs above, and the prompt architecture is carried by the system. If success depends on every user becoming good at prompting, adoption stops at the people who enjoy prompting.

3. Ground it in your own material

A general model gives general answers. Grounding the tool in Archetype context, meaning how the firm scopes work, how it writes and what it has delivered, is the difference between a novelty and something a consultant will put in front of a client. It also addresses the problem that started this: offices pasting client material into consumer chatbots, because a tool that already knows the firm's material is a better route than an improvised one.

4. Onboard leaders, not everyone

I trained thirty-plus senior leaders across twenty country offices first, deliberately. A tool introduced by your own regional lead tends to survive. A tool introduced by a group email tends not to. Adoption travelled through the org chart rather than around it. The groundwork was done before the tool arrives, so the people who will be asked about it already understand it.

5. Close the loop

What people actually ask the tool to do shapes the next iteration. The gap between the tool I designed and the tool people used was the most useful data in the project.

ABI, what is known and what is not
ClaimStatus
30+ senior leaders across 20 offices trained on AIDone
Internal demo surfaced 57 qualified opportunities in 15 minutes across 16 sectorsDemo result, not a production number
Design target of three to five qualified leads per day, per sector, per countryA target, not a result
Pilot, beta, workshop, launch (October 2026 to January 2027)Planned as of September 2026

What I would apply elsewhere

Three things travel to other organisations. Start from resented tasks, not from capabilities. Remove the need for skill at the point of use. Spend your early effort on the few people whose example others follow. What does not travel is the specific tooling, which depends on your data and your risk rules. The stack for ABI includes OpenAI, Gemini, Claude, Notion AI and Make.com models and tooling, chosen for this firm.

This post is updated only when the status above changes, and the date at the top moves with it.