What founders should know before investing in AI
AI can be useful, but only when it is tied to a real workflow, owned by the team, and measured against a business outcome.
AI is worth taking seriously. It is also very easy to waste money on.
The mistake I see most often is starting with the tool. A founder sees a demo, a competitor announces something, or a supplier suggests an AI feature. Suddenly the business is asking where AI can be added.
That is the wrong starting point.
The better question is: where does the business lose time, quality, margin, or decision speed today?
AI should be considered after that question has been answered, not before.
Start with the work
Every useful AI project starts with a real workflow.
How does a lead become a proposal? How does a support ticket become a resolution? How does a brief become a plan? How does a project manager produce a weekly update? How does a salesperson prepare for a call? How does finance chase missing information?
These workflows are usually full of manual effort. Some of that effort is valuable judgement. Some of it is admin, formatting, searching, rewriting, summarising, checking, and moving information between systems.
AI can help with the second group. It should be treated carefully around the first.
If you cannot describe the current workflow, you are not ready to automate it.
Do not buy theatre
There is a lot of AI theatre in the market.
Theatre is a chatbot on top of poor data. Theatre is a demo that works once but has no owner. Theatre is an automation that saves ten minutes and creates two hours of checking. Theatre is a vendor promising transformation without understanding how the business actually operates.
Useful AI is usually less dramatic.
It helps a team draft, classify, search, summarise, check, route, compare, or prepare. It removes friction from work that already exists. It gives people a better first version, a faster review path, or a cleaner decision surface.
That may sound less exciting than “AI transformation”, but it is where the value usually lives.
Ownership matters
An AI workflow needs an owner.
Someone must understand what it is meant to do, where the data comes from, what good output looks like, when a human must review it, and what happens when it fails.
Without ownership, the business gets a fragile tool. It may work for a while, but nobody trusts it enough to rely on it, and nobody understands it well enough to improve it.
This is why small pilots are useful. Not because the business should think small forever, but because pilots reveal whether the workflow, data, team, and operating model are ready.
The sensible test
Before funding an AI project, ask five questions:
- What business problem are we solving?
- Who owns the workflow after launch?
- What data does it need, and is that data reliable?
- Where does human review remain necessary?
- How will we know whether it worked?
If those questions cannot be answered in plain English, the project is not ready.
AI can absolutely make a business better. But it should be used with the same discipline as any other operational change.
The goal is not to look advanced. The goal is to remove friction, improve decisions, and leave the team with something it can trust.