Context matters

Do we really want everyone in the company to get the same answer?

At first glance it sounds sensible – one version of the truth, one interpretation, one message. In practice, however, an organization is not a homogeneous organism. It is made up of roles, accountabilities and perspectives that determine how we interpret the same piece of information.

One of the most common mistakes in designing AI assistants is assuming that a single answer to a question is entirely sufficient. It is not.

Take a simple example:
“Tell me how many days of leave I have?”

From an employee’s perspective it is a question about the available allowance and the chance to plan time off.
From HR’s perspective it is a matter of how the entitlement is calculated, the legal basis, the working-time fraction, length of service or changes in the regulations.
From a manager’s perspective – it is information about a team member’s availability, the impact on delivering targets and the need to reorganise work.

The same question. Three different contexts. Three different information needs.

PwC’s “Global Workforce Hopes and Fears Survey 2025” shows that employees expect digital tools to be personalised and information to be matched to their role. At the same time leaders point out that a lack of context-aware support in digital systems reduces the effectiveness of operational decisions.

Accenture’s “Technology Vision 2025” report, in turn, stresses that organizations achieving the highest productivity from AI design their systems around so-called role-based intelligence – intelligence assigned to a specific function in the company structure. Universal answer models turn out to be insufficient in an environment of complex processes.

Why?

Because a question inside an organization is never neutral. It is always “born” out of a role.

An employee asks because they want to plan something.
HR asks because they have to verify something.
A manager asks because they are accountable for the team’s results.

If an AI assistant ignores those differences and gives everyone an identical answer, it produces information that is formally correct but operationally inadequate.

And in an organization, adequacy is what counts.

Boston Consulting Group’s 2025 report on enterprise AI adoption indicates that one of the main factors increasing rollout effectiveness is matching systems to the structure of accountability. Companies that design AI with competence and decision-making differences in mind achieve productivity gains faster than those deploying universal solutions.

An assistant without context is a fast tool for generating content.
An assistant with context becomes decision support.

The difference is that in the second case the system understands:
– who the user is,
– what permissions they have,
– what they are accountable for,
– where they are in the process.

In an organization a question is never detached from the structure. Even the simplest query relates to a role, an accountability and a place in the hierarchy.

That is why designing an AI assistant should not start with “what answer should we give?”, but with “who is this answer for and what is it for?”.

One answer for everyone sounds like order.
In practice it is a simplification that ignores the real complexity of a company.

And complexity – properly understood – is a source of advantage, not a problem.