AI enablement

AI enablement is the work between buying AI and getting a useful result from it. The way I use the term, that includes making the right information available, changing the process, helping people learn the new way, and building small versions until you find what works.

You need a business result to aim for, otherwise you can finish the training and still not know whether the work improved.

What does AI enablement actually include?

I split it into four jobs, and in a small firm the owner needs to make sure all four happen. This is broader than sales enablement, which is about helping a sales team do its work.

Data design. A lot of how the work gets done lives in people’s heads, so the first job is making the records and rules AI needs available to it. That includes writing the process down before buying the agent.

Gallup found that, within organizations that make AI available, 88% of employees who strongly agree that AI fits their existing systems and processes use it frequently, compared with 55% of those who do not strongly agree. That’s an association in the survey, not proof that improving fit will produce the same change in your firm, but it’s a reason to look at how the tool fits the actual job.

Process redesign. You can put a faster tool into the old process and leave everything around it unchanged. But someone still needs to decide how review, handoffs, and the work assigned to the team should change if that time is freed.

Change management. People need hours to learn, written permission to use AI on real work, and a decision about what any freed time is for. Slack’s Fall 2024 Workforce Index found that 45% of desk workers had no explicit permission to use AI at work, and making that permission clear is something the owner can take responsibility for.

Prototyping. Build something small, put it in front of the person who would use it, and change it based on what breaks. That’s how you find out whether the output helps with the work, rather than only whether the software runs.

Is AI enablement the same as AI training?

No, because learning the tool is only part of changing how the job gets done. A training course can end when the sessions finish, but I wouldn’t call enablement finished without checking the business result we set out to change.

In an interview study of ten experienced U.S. professionals using M365 Copilot at work, no participant named formal training as their main way of learning it. Eight of the ten relied on trial and error and six on swapping tips with colleagues.

Ten people is a small interview study, not a rate you can apply to other firms, and it doesn’t show that training is unnecessary. It does give you a reason to make room for practice and help from colleagues rather than relying on a course alone.

Why can an AI enablement program stall?

The problem with selling readiness as the goal is that nobody knows what has to be true for the work to be over. Being prepared or equipped matters, but you can keep paying for those things without ever checking whether the business got a useful result.

Gallup found that 47% of U.S. employees say their organization has integrated AI, while 30% use it regularly. Gallup’s own summary is that implementation doesn’t guarantee use.

Those figures don’t explain why any one program stalled. They show why you need to check what happens after making the tool available, including whether using it helps the business.

What does AI enablement look like in practice?

I’d run it as a loop around one workflow and one result, rather than a set of phases that count as finished when the calendar says so. There are five steps to that loop.

Name the number first. One workflow, one measure that would change if this worked. A firm that cannot name it is not blocked on tooling, and finding that out costs nothing.

Write down what only lives in people’s heads for that workflow. Not the whole firm. The checks a manager runs without thinking, the exceptions, the conditions that make a case unusual.

Build the smallest version and put it in front of one person. Aim for something they can try on real work this week, with whatever checks that work needs, so you can learn from using it.

Make it easy to report that it broke. Ask what the person tried and what went wrong. That tells you what they wanted the tool to do, including things nobody wrote down, and what you need to fix next.

Change it and hand it back. Look at the information, the process, the support people have, and the build itself, then cut, rebuild, or try a smaller question. Keep checking whether those changes improve the result you named.

Nobody has a general formula for which build works, which is the honest reason this is a loop rather than a sequence of phases. The firms that get somewhere are the ones that can run the loop cheaply enough to run it several times.

Who owns AI enablement in a small firm?

You don’t usually need to start by hiring an enablement lead or forming a committee. In a small firm, the people changing the workflow may also be the people helping everyone learn it, so the owner needs to give them room to do both.

Look for the person already helping colleagues improve their workflows, writing down what works, and building small tools to remove repetitive work.

That person needs enough experience to help the people reporting problems, and people need to believe those reports lead to a fix. Otherwise, telling someone the tool failed can feel like admitting you couldn’t use it, and going back to the old way becomes the easier choice.

Don’t assume the tools you bought show you everything people use. Microsoft and LinkedIn’s 2024 Work Trend Index found that 78% of AI users bring their own tools to work, rising to 80% at small and medium-sized companies.

That’s a finding about AI users bringing tools, not a measure of why people stop reporting problems.

What does AI enablement cost?

The licence is only part of it, you also need to pay for the builds that don’t work and the time people spend trying them.

A prototype that doesn’t reach production can still tell you which data was missing, whether the problem was worth solving, whether anyone would use the output, and whether the results held up. Budget for more than one attempt, without treating another attempt as a promise that the next version will work.

The hours are the other real cost, and in a business they compete with billable work. That is a scheduling decision an owner has to make on purpose, which is the argument in nobody gave them the hours.

How do you know AI enablement worked?

Go back to the result you named before starting, such as the cost to deliver one service, the hours a recurring task takes, or a turnaround time clients complain about. Check whether it improved, including the time spent reviewing and fixing the work.

Seat counts, licence use, and course completions can tell you whether people have the tool and use it. They don’t tell you whether that business result improved, and if you can’t name the result yet, decide that before choosing another tool.

Where does AI enablement sit on the Delivery Model Ladder?

It is the work of moving from Stage 1 to Stage 2 on the Delivery Model Ladder.

Teaching people to prompt on top of the old process is Stage 1, Enhanced, where a task may get cheaper without changing the way delivery runs. At Stage 2, Augmented, you rebuild the workflow and decide what any freed time is for.

Enablement is the work between them, including paying for the prototypes and learning needed to change that workflow.

I describe the problems along the way in the Production Gap. The Owner Ceiling matters here because someone with authority has to make room for all four jobs, rather than expecting staff to want the tool enough to solve the process around it.

When is a firm not ready for this?

You need someone with authority to change the work, because enthusiasm for the tools doesn’t give a person permission to change other people’s jobs. You also need a result you can name, so the loop doesn’t become an open-ended readiness program.

And keep the size of the work honest. In a three-person firm, you may mostly be deciding what the owner does and what AI does, rather than coordinating a wider team, so don’t build a program larger than the problem.

The full argument, with what the loop cost to run in practice, is in AI enablement is a loop, not a roadmap.