From first AI baseline to building agents in four weeks

The challenge

A fast-growing beauty and wellness direct seller had the same challenge as almost every company right now. AI was everywhere in conversation and nowhere in the numbers.

Some leaders were experimenting on their own. One had already automated part of her workflow. Others used AI a few days a week for drafting and research. A few had barely started. Nobody knew where the team actually stood, which meant nobody could say what help each person needed.

The usual answer is company-wide training: the same slides for the person automating her reporting and the person who has never given AI more than a one-line prompt. It bores one and loses the other, and nothing changes.

The business had set itself a measurable goal instead: nearly half its leadership team at working proficiency with AI within months. To get there, it first needed to know the real starting point.

What we did

Inovara ran the programme in three connected parts as part of an AI Operating Partnership.

A measured baseline. Every leader sat a recorded one-to-one assessment: a structured, human conversation about how they actually use AI day to day, from prompting habits to how they check what comes back. Each transcript was then scored against a four-tier rubric (beginner, experimenter, practitioner, expert), with working proficiency set at 60 out of 100. Scoring every person against the same rubric made the results consistent and comparable, and gave the company a baseline it could trust.

The picture was clear. Across 17 leaders assessed, the average score was 36 out of 100. Nine landed in the beginner tier and eight were experimenters. Not one person had yet reached working proficiency.

A personal plan for every leader. Rather than a report that sits in a drawer, every score became a personal development plan: where that person is already strong, the three things to work on next and how they will know it is working. Each plan points to the specific sessions worth attending live, so nobody sits through content they have already outgrown.

Sessions built around real work. A weekly fluency programme runs alongside the plans, six sessions moving from foundations and prompting through judgement and checking to building and scaling. Every session is recorded, so the field never blocks the learning.

Then the part that makes it stick: hands-on working sessions with individual leaders and their teams, not on practice exercises but on the workflows they run every week. Scoping automations, building agents and putting AI to work on the problems people actually complain about.

Results

Four weeks in, the behaviour change arrived before the re-test did.

By the end of the first four weeks, 14 hands-on working sessions had run, one agent was live, two more were in build and six were scoped against named business problems. A department head with a team of 11 scoped five automations in a single working session, from inbox triage to automated monthly reporting. The first company hackathon ended with leaders presenting what they had built themselves, and generated enough energy that highlight reels were cut to bring the rest of the business in.

The baseline also surfaced capability the business could not see before. Its strongest profile had already cut days of feedback-board admin down to about an hour with an automation of her own, and her plan now focuses on turning those methods into documented patterns the rest of the cohort can borrow.

Just as telling was what the leaders started doing unprompted: booking their own sessions, sharing use cases with their teams and bringing new problems to the working sessions before anyone chased them.

The founder and chief executive summed it up in a voice note to the team: “You are exactly the kind of people that I love to work with.”

The cohort will be re-assessed on the same rubric at the end of the programme, with a target of 45% of leaders at working proficiency. The score lift gets measured then. The change in how the team works is already visible.

Why it matters

Most AI training starts with content and hopes for change. This programme started with a measurement.

A baseline turns AI capability from a feeling into a number, and a number can be managed. Personal plans mean every leader works on the gap that is actually theirs. And because the practice happens on live business problems rather than exercises, the learning pays for itself as it goes.

That is what building AI capability looks like when it is done deliberately: measure, personalise, build, re-measure.