Back to blogs

// What's on our mind

Spending AI budget on people before tools

A practical look at why AI budgets deliver more when coaching and AI training for employees get the larger share and tools get the smaller one.

Spending AI budget on people before tools

You'll get more from an AI budget when most of it pays for coaching people and far less pays for licences.

Most firms do the opposite. They sign the platform contract first, then leave a sliver for a lunch session and hope new habits follow. The gap shows up in the numbers. Only 15.9% of employed respondents told a November 2025 Federal Reserve Bank of New York survey that their employer currently offers any AI training. Tools are arriving far faster than the skill to use them.

That imbalance is why AI spending stalls at banks, insurers and wealth firms. A licence gives a claims adjuster or a credit analyst access. It doesn't show them which part of their Tuesday it should replace. We think the budget should follow the hard part, which is getting people to work differently and safely inside your controls. Fund AI training for employees first, and the tooling decision gets smaller, cheaper and easier to defend.

Why most AI budgets buy licences that nobody uses well

AI budgets lean toward tools because tools are easy to buy, easy to count and easy to report upward. Coaching is slower to scope and harder to show on a slide. So the money lands on seats and platforms, and the work of teaching people how to use them goes unfunded.

Picture a mid-sized insurer that rolls out an assistant to every employee in one quarter. Usage spikes for 2 weeks, then settles to a small group who were already curious. Underwriters try it on a submission, get a vague answer and go back to the old template. Nobody showed them how to feed it the policy wording or check its output. The licence bill keeps arriving every month anyway.

The cost of that pattern is more than wasted seats. Leaders read low usage as proof the technology was oversold, and the next funding request gets harder. Staff read it as one more tool pushed at them without help. You end up with a checked box on a board report and very little different about how work gets done.

What a people-first AI budget looks like in practice

A people-first AI budget treats the tool as the small line and the coaching as the large one. You buy a sensible platform, then spend most of the money putting experienced people beside your teams until daily work looks different. The exact ratio matters less than the order of priorities.

Try a thought experiment with $10 million. Spend roughly $1 million on tooling and choose something solid, because the platform is only the entry ticket. Put the remaining $9 million into coaches who sit with finance, legal, HR and sales every day. They show a controller how to reconcile faster and a lawyer how to review a contract clause with a second set of eyes.

That split sounds extreme until you look at where value appears. It appears when a person in a specific role finishes a specific task better than last month. Staff also keep the skill, which makes them more confident about what AI is doing around them. We'd rather see you fund that confidence than a second platform you don't need yet.

‍

A people-first AI budgetIn a $10 million thought experiment, roughly $1 million goes to tooling and the remaining $9 million pays for coaches who sit with teams every day.A $10 million AI budget$1Mtooling$9M coaches beside finance, HR, legal and salesThe platform is only the entry ticket, so most of the money pays for coaching.

‍

How AI training for employees turns licences into measurable output

AI training for employees pays back when it's tied to a live task, taught in the flow of work and repeated until it sticks. Generic courses teach what a prompt is. Role-based coaching teaches a loan officer to prepare a credit memo in fewer steps, with every check intact.

‍

"It appears when a person in a specific role finishes a specific task better than last month."

‍

Skills are the constraint employers name most often. Skill gaps rank as the biggest barrier to business change, with 63% of employers calling them a major barrier for the 2025 to 2030 period in a 2025 global employer survey. Consider a wealth operations team that handles account transfers. A coach spends 3 weeks mapping the steps, builds prompts for the exception cases and stays while the team runs live work.

The output is something you can count. Fewer transfers bounce back, handling time drops and the team knows why each check exists. That's a different result from a completion certificate. It also gives risk and compliance a clear view of how the tool is used, because they helped design the workflow.

‍

Skill gaps as the main barrierIn a 2025 global employer survey, 63% of employers called skill gaps a major barrier to business change for the 2025 to 2030 period.63%of employers call skill gaps a majorbarrier to change for 2025 to 2030Skill gaps rank as the biggest barrier to business change, named by 63% of employers.

‍

Where AI upskilling budgets go wrong inside regulated firms

AI upskilling fails in regulated firms when it's run as a single event, aimed at everyone equally and separated from governance. People leave a workshop with enthusiasm and no permission. Without clear rules on data, review and accountability, the safest choice for any employee is to stop using the tool.

Five mistakes drain these budgets faster than any others.

  • Training happens once, with no follow-up when the hard questions surface.
  • Content stays generic, so nobody sees their own job in it.
  • Managers get skipped, even though they set the tone for daily habits.
  • Data and privacy rules arrive after people have already improvised.
  • Success gets measured by attendance instead of work that looks different.

Take a bank that trains several hundred analysts in a single week. A month later, many of them aren't sure which client data they're allowed to paste into the assistant. Some avoid it entirely, and others guess. Both outcomes are a governance problem, and both trace back to how the money was spent. Coaching that includes your risk partners from day 1 closes that gap before it becomes an audit finding.

How to split spending between enablement work and new tools

The main difference between spending on enablement and spending on tools is what you own afterward. Tool spending buys access that ends when the contract does. Enablement spending builds habits, judgment and workflows that stay with your people and carry over to the next platform you adopt.

A simple way to budget for AI is to fund the smallest tool set that covers your priority use cases, then direct the larger share to coaching. Start with 2 or 3 workflows tied to a business goal you've already committed to. Each choice plays out differently across 5 budget questions.

‍

Budget questionTool-first spendingPeople-first spending
Where does the money go first?Licences and platform fees take most of the budget before use cases are known.A modest share covers one solid platform and the larger share pays for coaches.
How do staff learn the tool?People receive a launch email and an optional course they finish alone.People practise on their own live work with an experienced coach beside them.
When does risk get involved?Guidance on data and review shows up once staff have built their own workarounds.Risk and compliance partners shape each workflow before it goes live.
What gets reported upward?Reports count seats purchased and monthly logins.Reports track cycle time, error rates and manual tasks that were retired.
What is left after the contract ends?Access disappears and working habits look the same as before.Skills and redesigned workflows carry over to whichever platform comes next.

‍

Expect pushback on the ratio, because coaching looks like an operating cost while software looks like an asset. The answer is to tie every coaching dollar to a named workflow and a number you'll report. That keeps the spend right-sized to the problem and easy to defend at budget review.

Which measures prove coaching is paying back the budget

Coaching pays back when the work itself looks different, so measure the work. Track cycle time, rework, error rates and the number of tasks a team has stopped doing by hand. Set those numbers before the first session, then review them every month alongside adoption.

A payments operations team makes a good test. Before coaching starts, record how long a dispute takes to resolve and how often it gets reopened. Check the same 2 numbers at 30, 60 and 90 days. If nothing has moved, the coaching is aimed at the wrong task and you'll know early enough to redirect it.

We run engagements this way at Electric Mind. Our engineers sit inside client delivery teams and stay until the tools have altered how people do the job, and we treat that as the result worth reporting. Seat counts tell you what you bought. Workflow numbers tell you what you got, and they're the figures a CFO will keep funding.

Why patient coaching outlasts every tool you'll buy this year

The tools will keep being replaced, and the people who learned to use them well will still be on your payroll. That's the plain case for weighting an AI budget toward coaching. Skill compounds across every platform you adopt, while a licence only holds its value until the next renewal.

‍

"Seat counts tell you what you bought."

‍

We've watched new technology arrive many times, and the pattern holds. Firms that rush to check the box end up with a list of things they own. Firms that go slow at the start, set the principles and teach their people end up moving faster a year later. It isn't glamorous work. It's a coach showing someone in finance a better way to close the month, then coming back the next day to do it again.

So start with what you're trying to accomplish, pick the tool that fits and put the weight of the budget on people. That's the approach we take at Electric Mind, working with teams until the tools alter how they work. Your staff will come out of it able to say they've used AI and understand it. That confidence is what your budget should be buying.

Got a complex challenge? Let's solve it together, and for real.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Top