Notes

AI has to belong somewhere.

Engineer-led AI consulting team mapping high-friction workflows

AI is not useful just because it can do something. It becomes useful when it has a clear place in the work.

The AI bit is not enough

The question is not just whether AI can do something.

Most of the time, it can do something.

It can summarise, draft, search, compare, organise, rewrite, classify, suggest, and generate. The list keeps getting longer, and some of it is genuinely useful.

But that does not automatically make it a product.

And it does not automatically make the work better.

The more interesting question is where the AI actually belongs.

Output still needs somewhere to go

A lot of AI ideas start with the output.

A summary.
A draft.
A recommendation.
A chatbot answer.
A generated report.
A cleaner version of messy information.

That can all be useful.

But only if the output has somewhere useful to go.

Who is using it? What are they trying to do next? Does it help them make a decision, move work forward, reduce a repeated task, or understand something faster?

Or does it just create another thing to read, check, copy, move, or fix?

That is the difference I keep coming back to.

AI output on its own can look impressive. But if it does not fit into the work properly, it can become another layer instead of removing one.

Another tool is not always progress

This is where I think some AI products can get messy.

The AI works, technically. It produces something. It responds. It generates. It does the thing you asked it to do.

But then what?

If someone has to take that output, copy it somewhere else, rewrite it, check it against another system, send it to someone manually, and then remember the next step, the AI may have helped a little.

But the workflow is still doing a lot of the damage.

That does not mean the AI is useless. It just means the product thinking cannot stop at the AI bit.

Where it sits matters.

The boring placement questions

This is the less exciting part, but probably the more useful one.

Where should the AI appear in the process?

Before someone starts the work?
While they are doing it?
After they have created something?
Only when something is missing?
Only when the system is unsure?
Only when a human needs a first draft, not a final answer?

These are not glamorous questions.

But they matter because AI in the wrong place can slow things down. It can create more checking, more uncertainty, more steps, and more noise.

AI in the right place can do the opposite.

It can pull context together before someone starts. It can turn scattered notes into something usable. It can help spot what changed. It can suggest the next step. It can make the work feel less heavy without pretending to replace the person doing it.

That feels like a better goal.

Useful AI often feels less dramatic

The useful version is not always the most impressive from the outside.

Sometimes it is just a small part of the workflow that gets easier.

A user does not have to search through five documents.
A founder does not have to rewrite the same update again.
A team does not have to keep asking where the latest version is.
Someone gets a useful first draft instead of starting from nothing.
A messy set of information becomes easier to understand.

That is not always exciting to show in a demo.

But it is closer to real value.

Because the point is not that AI exists inside the product. The point is that it helps the work move.

It still needs judgement

This is also why I do not think “just add AI” is a very useful way to think.

AI has to have a job.

Not a vague job like “make this smarter”.

A real job.

Reduce this repeated step.
Help someone understand this faster.
Draft this so they do not start from scratch.
Check this before it moves forward.
Pull this context together.
Make this decision easier, without pretending to make the decision for them.

That is when AI starts to feel useful.

Not because it is magical.

Because it belongs somewhere.

The question I would ask

So when looking at an AI idea, I would not only ask:

Can AI do this?

I would ask:

Where does this belong in the work?

That question changes things.

It makes you look at the real task, the user, the workflow, the handover, the next step, and the places where human attention is being wasted.

It also makes the idea harder to fake.

Because a clever AI feature can sound good in isolation. But if it does not have a clear place in the work, it probably will not matter much for long.

AI has to belong somewhere.

That is where it starts becoming useful.

Linda Eriksson

I write about AI, automation, and the practical work of turning ideas into useful software.

Start the conversation

Big plans. Let’s bring them to life.

Turn your next idea into software your business can rely on, with a technical team that takes ownership from day one.

Let’s talk about your project