AI doesn’t need more prompts. It needs business context.
4 Sept 2026 | Amanda Cunningham | Founder, Savvy Pixel®
AI doesn’t need more prompts. It needs more business context.
I’ve been thinking a lot lately about how businesses are actually using AI day to day.
For the last couple of years, so much of the conversation has been about prompting.
How to write a better prompt. How to get a better answer. How to build a prompt library.
And prompting does matter.
But I’m increasingly wondering whether we sometimes put too much pressure on the prompt itself.
Because even a very good prompt can only go so far if the AI doesn’t really understand the business behind it.
What are we trying to achieve?
Who are our customers?
How do we normally communicate?
What matters commercially?
What information should never be used?
Where does someone need to review the output before it goes any further?
Those things aren’t really prompting questions.
They’re business-context questions.
We’re all still working this out
I don’t think there is a business anywhere that has completely solved AI adoption.
Most organisations are experimenting.
Someone is using ChatGPT for emails.
Someone else is trying Claude for research.
Another person has discovered something useful in Gemini.
A colleague has built their own way of working and is getting great results.
That is often exactly how adoption starts.
The challenge comes when individual experimentation begins to spread across a business.
One person gives AI plenty of context.
Another assumes it already knows.
One person checks everything carefully.
Another trusts the first answer.
Someone understands what information is appropriate to share.
Someone else may never have been given that guidance.
None of this necessarily means people are doing anything wrong.
It usually means the organisation has moved into AI faster than its shared working practices have developed.
And that feels like quite a normal stage to be in right now.
What does that look like in the real world?
Sometimes it’s surprisingly ordinary.
Imagine a sales team using AI to prepare for customer meetings.
One person tells the AI who the customer is, what they already buy, the history of the relationship and the commercial objective for the meeting.
Another simply asks:
“Give me some questions to ask this customer.”
Both are using AI.
But they are unlikely to get equally useful results.
Or take marketing.
A team asks AI to draft content, but the AI hasn’t been given a clear picture of the audience, the proposition, tone of voice, evidence it can rely on or claims it should avoid.
The first draft might look polished.
But somebody still has to spend time making it sound like the organisation, checking what it has assumed and correcting anything that isn’t quite right.
In customer service, the issue might be different.
AI can make drafting a response much faster.
But what happens when a complaint involves a refund, a vulnerable customer, confidential information or a decision that should be made by a person?
The useful question isn’t only:
“Can AI write this response?”
It is also:
“Where should AI stop?”
Or consider a manager writing a proposal or report.
AI might be very good at structuring the document.
But if it doesn’t understand the organisation’s priorities, previous decisions, terminology and commercial constraints, the manager can end up repeatedly supplying the same background information every time.
None of those examples is particularly dramatic.
That is the point.
AI adoption is increasingly happening inside ordinary work.
And that is why context, boundaries and shared working practices matter.
The interesting question is what comes next
Perhaps the next phase is less about teaching everyone to become brilliant prompt writers and more about creating a better shared foundation.
What should our AI systems understand about the business?
Which tasks are genuinely useful applications of AI?
What context should staff have available?
What should remain private?
Where might accuracy matter more than speed?
When should AI use be transparent?
And where does accountable review need to sit?
Once some of those questions are clearer, prompting becomes easier too.
People are no longer starting from a blank page every time.
Context can remove a surprising amount of friction
We probably all recognise some version of this.
You open a new AI conversation and spend the first few minutes explaining who you are.
Then you explain the business.
Then the customer.
Then the tone.
Then what you definitely don’t want.
Then you correct the first answer because it sounds nothing like you.
That might be manageable for one person.
Across a team, though, that repeated explanation can become a lot of duplicated effort.
And it can lead to very different outputs depending on who happens to be using the AI.
That is why I think business context is becoming so important.
Not because we need another complicated AI framework.
Quite the opposite.
It can simply mean giving people clearer shared information and boundaries so they’re not reinventing the relationship with AI every time they use it.
What if we treated AI like a new colleague?
There’s another way I’ve started to think about this.
If someone joined your organisation tomorrow, you probably wouldn’t hand them a laptop and say:
“Off you go. Work it out.”
You would introduce them to the business.
You would explain what the organisation does.
Who the customers are.
How the team works.
What good looks like.
What they’re responsible for.
What they’re not authorised to do.
Where to find important information.
And when they need to involve somebody else.
In other words, you would induct them.
So why do we so often expect AI to work effectively inside a business without giving it anything resembling the same foundation?
That is where the idea of AI Induction comes in.
Not induction in the sense of teaching people how to use a chatbot.
Induction for the AI itself — and for the way the organisation works with it.
AI Induction as an operating system
I’m increasingly seeing AI Induction as the starting point for an organisation’s AI operating system.
Not a piece of software.
And not another layer of technology.
An operating system in the practical business sense: the shared context, rules and working practices that sit underneath everyday AI use.
That might include:
Business context What the organisation does, its priorities, customers, proposition, terminology and ways of working.
Useful applications Where AI can genuinely reduce effort, improve thinking or support better work.
Clear boundaries Privacy, confidential information, accuracy, transparency and appropriate use.
Accountable review Where somebody needs to check, approve or take responsibility for an output or decision.
Working environments How useful context is retained in Projects, workspaces, Gems or other appropriate AI environments instead of being explained from scratch each time.
Role context Giving different people the information and boundaries relevant to the work they actually do.
Suddenly the conversation becomes much less about:
“What prompts should we give everyone?”
And much more about:
“How do we want AI to work inside this organisation?”
For me, that is a much more useful question.
This is also a commercial conversation
The part I find particularly interesting is that this isn’t really just about technology.
It affects ordinary business questions.
Time.
Consistency.
Customer experience.
Knowledge.
Risk.
Decision-making.
If people spend less time repeatedly explaining the business, that is useful.
If AI outputs are more relevant first time, that is useful.
If colleagues have clearer boundaries around privacy and accuracy, that is useful.
If organisational knowledge is less dependent on one person remembering everything, that is useful too.
And if five people can use AI in ways that are appropriate to their different roles while still working from the same organisational foundations, that starts to look much more like capability than experimentation.
So perhaps the value of better AI adoption isn’t simply “doing more with AI”.
It may be reducing some of the friction already sitting inside the way we work.
Foundations first
AI platforms will keep changing.
There will be new features, new models and plenty of claims about what businesses should be doing next.
We don’t need to respond to every one of them.
But having stronger foundations makes those developments much easier to judge.
Does this actually improve something for our business?
Is it appropriate for the work?
Do we understand the implications?
And can we use it with clearer boundaries around privacy, accuracy, transparency and accountable review?
Those feel like increasingly useful questions to keep asking.
Because becoming AI-capable probably won’t mean using AI everywhere.
It will mean understanding where it adds value, giving it the right context and keeping people accountable for the decisions that matter.
That is the thinking behind the Savvy Pixel® AI Induction Hub: a practical starting environment for organisations that want to move from individual AI experimentation towards a more consistent way of working.
Not simply:
“Here’s how to use AI.”
But:
“Here’s how AI should work within our business.”
And I suspect that distinction is going to become increasingly important.
I’m Amanda Cunningham, founder of Savvy Pixel®.
I work with organisations to bring their digital picture into focus and build clearer foundations for practical, responsible AI adoption.
Savvy Pixel® — Clearer digital decisions for the AI era.
Originally published through Amanda Cunningham’s LinkedIn newsletter on 4 September 2026.
