Project 02AI & Enablement
ABI
An internal AI tool for a twenty-office consultancy, and the enablement programme built to make it survive contact with people who have deadlines.
Pre-pilot, as of September 2026
- 20country offices
- 30+leaders onboarded
- Buildlead
01
At a glance
The whole project in three steps.
- The problem
Twenty offices were running twenty separate AI experiments, and nothing one country learned reached the next.
- What I did
Designed and led the build of an internal AI tool, and trained 30+ leaders before it arrives rather than after.
- The result
Pre-pilot. An internal demo surfaced 57 qualified opportunities in 15 minutes; the pilot was scheduled for October 2026.
Skills it shows
- AI enablement
- Internal tool design
- Stakeholder enablement
- Process mapping
- Adoption planning
Tools
- OpenAI
- Gemini AI
- Claude AI
- Notion AI
- Make.com
02
The challenge
Why the project had to happen.
Generative AI arrived at Archetype the way it arrived everywhere: as twenty separate experiments. Individual offices were pasting client material into consumer chatbots, prompts lived in private notes, and nothing anyone learned in one country reached the next.
Read the full context
The brief was not "adopt AI". Every firm says that, and most of them mean a pilot that quietly ends. The brief was to make it survive contact with people who have deadlines. That is a product problem and a behaviour problem at the same time, and the second one is harder.
03
How I did it
5 steps. Open any one for the detail.
Find where the time actually goes
Mapped the repeated work first: proposals, scope translation, reformatting for twenty markets.
How
Before building anything, I mapped the repeated work: proposal boilerplate, translating technical scope into client language, reformatting the same content for twenty markets. The tool had to attack tasks people already resented, or it would lose to habit.
Build for the reluctant user
Designed for the reluctant user: structured entry points, no prompt writing needed.
How
ABI was designed around the person who does not want to learn a new tool. No prompt engineering required at the point of use: structured entry points for the jobs above, with the prompt architecture carried by the system rather than the user.
Ground it in our own material
Grounded the model in how Archetype scopes, writes and what it has delivered.
How
A general model gives general answers. Grounding it in Archetype context (how we scope, how we write, what we have delivered) is the difference between a novelty and something a consultant will put in front of a client.
Onboard leaders, not everyone
Trained 30+ leaders first, so adoption travelled through the org chart.
How
Thirty-plus leaders first, deliberately. A tool introduced by your own regional lead survives; a tool introduced by a group email does not. Adoption travelled through the org chart rather than around it.
Close the loop
Let what people actually asked for shape the next iteration.
How
What people actually asked for shaped the next iteration. The gap between the tool I designed and the tool people used was the most useful data in the project.
04
See it
A rebuild, not the product. ABI is an internal Archetype tool, so nothing below is lifted from it. The interaction model is the one this case study argues for, rebuilt with invented content. No real client, project or figure appears in it.
ABI — reconstruction
Demo — scripted, no live model
Every field, client name and output below is invented. The interaction model is the one described above; the content is not lifted from the real tool.
Pick a task and press Draft.
Rollout — twenty country offices
- Vietnam
- Cambodia
- Laos
- Myanmar
- Thailand
- Philippines
- Indonesia
- Malaysia
- Singapore
- Mongolia
- India
- Sri Lanka
- Bangladesh
- UAE
- Qatar
- France
- Poland
- Romania
- Kenya
- Peru
05
What it achieved
An internal demo surfaced fifty-seven qualified opportunities in fifteen minutes across sixteen sectors. That is a demo result, not a production number, and it is what moved the project from argument to roadmap.
Thirty-plus senior leaders across twenty country offices have been trained on AI and given structured entry points, so the enablement groundwork is done before the tool arrives rather than after.
Pre-pilot as of September 2026: pilot October, beta November, business-development workshop December, launch January 2027. The design target is three to five qualified leads per day, per sector, per country.
06
What I took from it
The gap between the tool I designed and the tool people used was the most useful data in the project.
A tool introduced by your own regional lead survives. One introduced by a group email does not, so I onboarded leaders first.
A tool has to take on tasks people already resent, or it loses to habit.