What Should Leadership Actually Do About AI This Quarter?

By Claire Bicong, Co-Founder · 16 August 2026

Most leadership teams have had the AI conversation four or five times by now and still have nothing on the roadmap. Someone shares an article, someone demos a tool, everyone agrees it matters, and the meeting ends. A quarter later the same conversation happens again with a different tool.

Meanwhile AI is already in the building. Marketing has a chatbot open all day, someone in finance is pasting figures into a free tool, and a manager signed up for something on a company card. The risk for most companies right now is not that AI arrives too fast. It is that it arrives with no direction, no rules, and no owner.

This is the ground we cover in our executive and leadership briefings. Below is the short version: what leadership actually has to decide, and what a realistic first 90 days looks like.

Why do executive AI conversations go nowhere?

Because they start with tools. Someone asks which model is best and the meeting turns into a product comparison nobody in the room can settle, on a market that changes again before the next board pack. The answer keeps moving, so the decision never lands.

The useful question is narrower and it holds still: which repeated work in this business costs us the most hours, and how much of that work is writing, summarizing, sorting, or drafting? That is answerable in an afternoon with your own team, and it points at a decision instead of a debate.

Where does AI create real leverage in a business?

In three places, roughly in the order they pay back. Work that repeats every week. Work that is mostly text. Work that queues up behind one busy person.

A weekly client report that takes four hours is worth more than a clever one-off use case, because you collect the saving 52 times. When we rebuilt the systems and documentation flow for Wood's, the Swedish indoor climate brand, the win was not a new capability. It was 20 hours a week going back to the team that was drowning in admin.

The mistake we see most often at leadership level is chasing the impressive use case instead of the frequent one. Frequency is where the money is.

What does leadership have to decide?

Four things, and none of them require you to understand how a model works. Make these calls and your teams can move without asking permission at every step.

• Approved tools: which AI tools are allowed, on company accounts, paid where it matters. Free consumer accounts are where company data quietly goes to train someone else's model.
• The data rule: what can never be pasted into an AI tool. Client data, employee data, anything under contract or under GDPR. One page, written down, in plain language.
• An owner: one named person accountable for the rollout. Not a committee. AI adoption dies in committees.
• A 90-day definition of good: hours saved on a named process, or more output at the same quality bar. If you cannot name the number now, you will not know later whether it worked.

Should you build or buy?

Buy first, almost always. Off-the-shelf tools plus well-built workflows cover the majority of what a small or mid-sized company needs, and they cost a few hundred euros a month instead of a development budget.

Build only after a workflow has run manually long enough that you know exactly what it does, how often it runs, and what it is worth. Custom AI projects that skip that step tend to automate a process nobody had agreed on yet, which is an expensive way to discover your own operations.

How do you roll it out so it actually sticks?

One department at a time, starting with the one that has the most repeatable text work. Usually that is marketing or operations. A company-wide AI day feels decisive and changes very little, because general training gives everyone the same generic examples and nobody leaves with their own job made easier.

Department-specific training works because the team spends the session on their own tasks, with their own documents, and walks out with real work already done. That is why we split our training into department tracks for marketing, operations, sales, HR, and leadership rather than running one course for everybody.

What does a realistic 90-day plan look like?

• Days 1 to 30: pick the tools, write the data rule, name the owner. Run one department through proper training and choose two processes to change.
• Days 31 to 60: measure those two processes against what they cost before. Fix what broke. Write down the prompts and workflows that worked so they survive the person who built them.
• Days 61 to 90: take the proven workflows to a second department. Review the numbers with your owner and decide what gets budget next quarter.

That is a quarter of real progress with no custom build, no transformation programme, and no consultant living in your office.

The short version

Stop comparing tools and start counting repeated work. Approve the tools, write the data rule, name an owner, and set a number you want to hit in 90 days. Train one department properly instead of everyone vaguely. Buy before you build. The companies pulling ahead on AI are rarely the ones with the best model. They are the ones who made four decisions early and let their teams get on with it.

Our executive AI briefing is a two-hour working session that produces exactly that: the leverage map for your business, the guardrails, and a 90-day plan with owners attached. Programs run €1,400 to €4,000 depending on format and scope. If you want the foundations first, our free AI modules are open to everyone.

Claire Bicong is co-founder of bodioagency, a marketing, operations, and AI training agency in Valencia, Spain. She has spent 13+ years in fintech, AI, and consumer brands building growth and operations systems that run without heroics.