Your organisation is not built for AI. Here is how to change it.

You're reading The AI Directive — practical AI intelligence for
business leaders. Subscribe
here if someone forwarded this.

Most organisations now have access to broadly similar AI
capabilities.

They can summarise documents, analyse data, draft content, compare
options, automate routine work and put increasingly capable agents
around business processes.

That access matters. But it will not create lasting advantage on its
own.

The real difference will come from what the organisation is capable
of doing with it.

Can people redesign work rather than simply accelerate their existing
tasks? Can decisions move at the speed of better information? Is the
data reliable enough for machines and humans to act on? Can technology
connect to the flow of work rather than sit beside it as another
application?

Those are not primarily tool questions.

They are questions about how the organisation operates.

In #018, I argued that faster individual work does not automatically
create a faster business. In #019, we saw how better analysis can simply
expose slow decisions and unclear authority.

The positive conclusion is not that AI has failed.

It is that leaders now have a much bigger opportunity: to redesign
the organisation around what AI makes possible.

Many AI programmes begin by collecting use cases.

That is a reasonable way to create energy and learn quickly. Teams
identify repetitive tasks, trial tools and demonstrate time savings.
Useful ideas emerge.

But a list of use cases does not add up to a new operating model.

It often leaves the underlying work unchanged. One person drafts
faster, but the same review chain remains. A report is produced
automatically, but managers still wait for the monthly meeting. Customer
information is summarised instantly, but nobody owns the decision it
should trigger.

The better starting point is a business outcome.

Choose something that matters: reducing the time to launch a product,
improving customer retention, increasing first-time-right quality,
accelerating a commercial decision or resolving service failures before
customers chase.

Then ask how people, process, data and technology must work together
to improve that outcome.

That shift sounds small. It is the difference between deploying a
capability and transforming a system.

Giving people licences is not the same as enabling them to change the
business.

The organisations that benefit most will create AI
agency: the confidence, permission and capability to examine
how work is done, challenge unnecessary steps and build better ways of
operating.

That requires more than prompt training.

People need enough understanding to judge where AI is reliable, where
it needs supervision and where human context remains essential. They
need access to the right tools and data. They also need clear boundaries
within which they can experiment and improve work without navigating a
project approval process for every sensible change.

Managers have a particularly important role.

As AI reduces information gathering and routine coordination, good
management becomes more—not less—valuable. Managers create context,
frame choices, coach judgement, resolve competing priorities and notice
when local optimisation damages the wider outcome.

Their job should move away from carrying information up the hierarchy
and towards helping people and AI act intelligently closer to the
work.

Leaders can enable this by doing three things:

The aim is not an organisation in which everyone uses AI every
hour.

It is one in which people can recognise and act on the opportunities
that matter.

AI creates the most value when it changes the flow of work.

That means looking beyond the task it performs to what happens before
and after it.

If AI can prepare a customer proposal in minutes, does the commercial
process still need the same sequence of handoffs? If it can monitor
operational signals continuously, should management wait for a weekly
report? If it can assemble the evidence for a routine decision, can
authority move closer to the person responsible for the outcome?

The strongest AI-enabled processes share three characteristics.

First, they are designed around the end-to-end outcome rather than
the convenience of individual functions.

Second, routine work moves with minimal intervention while genuine
exceptions are surfaced to people with the context needed to act.

Third, they include feedback. The system does not merely execute; it
shows what happened, where judgement was needed and how the process
should improve.

This is where AI and established operational improvement disciplines
belong together. Flow, constraints, standard work, visual management and
exception handling are not old manufacturing ideas made irrelevant by
AI. They are precisely the disciplines that stop powerful technology
being absorbed into a weak process.

The question is not, “Where can we insert AI?”

It is, “How should this work now flow?”

AI exposes the quality of an organisation's data very quickly.

It also tempts leaders into believing that the answer is another vast
data programme.

Sometimes foundational investment is unavoidable. But the practical
route is to connect data improvement to the decisions and outcomes the
organisation is trying to change.

For each priority flow, identify:

This turns data from a technical estate into a decision asset.

It also helps avoid a common failure: using AI to create polished
answers from inconsistent inputs. Fluency can disguise uncertainty. A
well-written recommendation assembled from disputed measures is still a
weak recommendation.

The positive opportunity is significant. When reliable operational
and customer data is available in the flow of work, AI can make patterns
visible earlier, tailor information to the decision and reduce the time
people spend assembling evidence.

Good data does not replace judgement.

It allows judgement to be used where it adds most value.

The first wave of enterprise AI has largely arrived through
general-purpose assistants.

That has been valuable. It has given people direct experience and
revealed thousands of small productivity opportunities.

The next step is integration.

AI needs to connect safely to the systems, knowledge and events that
make up real work. It should be able to recognise when something has
changed, assemble the relevant context, support or execute the permitted
action and leave a clear record of what happened.

This does not require a heroic attempt to rebuild the whole
technology estate around AI.

It requires a deliberate architecture:

Technology then becomes part of the operating system of the business
rather than a collection of clever demonstrations.

The distinction matters. A chatbot can answer a question. An
AI-enabled organisation can recognise what needs attention, bring
together the evidence, route the decision to the right place and learn
from the result.

People, process, data and technology are often presented as four
workstreams.

That is tidy—and dangerous.

The value sits in the connections between them.

A team cannot take more ownership if the process still requires
approval three levels above. A redesigned process cannot operate
dynamically if the data arrives monthly. Better data has limited value
if technology cannot surface it where the decision happens. Connected
technology will disappoint if people do not trust it or understand the
boundaries of their authority.

This is why AI transformation cannot be delegated entirely to IT,
data, HR or an innovation team.

Each function has essential expertise. None owns the whole
outcome.

Senior leadership must establish the direction, choose the priority
outcomes and resolve the cross-functional trade-offs. The work itself
must then be redesigned with the people who understand how it really
operates—not only how the process documentation says it operates.

The organisation chart may eventually change as a result. Roles may
combine. Management spans may alter. Some activity will disappear and
new capability will be required.

But structure should follow the redesigned work.

It should not substitute for it.

Do not begin with an enterprise reorganisation. Choose one meaningful
cross-functional outcome and prove the new way of working.

Define the outcome and its current measures. Map how work,
information and decisions flow today. Identify the constraint, the
avoidable handoffs and the points where human judgement genuinely
matters.

Assess the four elements together:

Create a simpler flow. Give named people clear decision boundaries.
Use AI to remove or compress specific work, not merely to produce more
output. Put the minimum reliable data into the process and test with
real cases.

Keep the scope controlled enough to learn safely, but material enough
that the result matters.

Compare the new outcome with the baseline. Measure elapsed time,
quality, customer impact, cost, decision speed and the volume of
exceptions requiring human intervention.

Capture what changed across all four elements. Then decide whether to
extend the pattern, adjust it or stop.

The deliverable is not another AI use case.

It is a repeatable method for redesigning how the organisation
works.

Choose one cross-functional outcome worth improving over the next 90
days.

Name the executive owner, establish the baseline and put people,
process, data and technology around the same table.

Do not ask each group for its AI plan.

Ask them to design one better system together.

This arc began with a paradox: AI had made work faster, but the
business was still slow.

The answer was never to ask people to work even faster.

It was to redesign the system around them.

AI gives leaders an opportunity to build organisations in which
better information leads to better decisions, routine work flows with
less friction and people spend more of their time on judgement,
relationships and improvement.

That is a far more valuable ambition than automating yesterday's
organisation.

— Toby

@TobySmithFirth on
YouTube

Subscribe to The AI Directive here: https://theaidirective.co.uk/subscribe?utm_source=beehiiv&utm_medium=email&utm_campaign=issue_020

These are tools I use personally. Affiliate links marked - I earn a small commission if you sign up, at no extra cost to you.

Some links in this issue are affiliate links. I only recommend tools I actually use.