For the past 18 months, most conversations about AI in insurance have revolved around adoption. Now we need to focus on implementing it.

By Tim Mole, Head of Document Intelligence, JAVLN (pictured)

From my perspective, insurance broking is not short on interest in AI.

What it has been short on is the right foundation. Many firms are still carrying an admin tax created by disconnected systems, repeated data entry and workflows that do not reflect how brokers actually work.

Workflow friction is still the core issue

When people talk about inefficiency in broking, the assumption is often that the broker management system itself is the problem. In many cases, it is not that simple. The bigger issue is the friction between systems.

Brokers are often re-entering the same client and risk information into multiple places. They move between a broker management system, a CRM, insurer portals, file notes, document repositories and email. Each extra click may look minor in isolation, but across a renewal cycle or a working week, it becomes substantial.

This is where the admin tax becomes debilitating.

Time gets absorbed by administration instead of being spent where brokers make the biggest difference: advising clients, strengthening relationships and growing their book.

Legacy technology is still shaping modern work

The industry is also dealing with a simple reality. Some of the technology brokers rely on is 30 years old. That often means more steps, more clicks and more manual workarounds than anyone would design today.

It also creates a mismatch between the way brokers work and the way systems expect them to work. Brokers are in meetings, on the road and in constant contact with clients.

Yet many workflows still assume they are sitting at a desk, gathering information in one moment and entering it properly later.

That gap matters. It is one reason mobility, simplification and automation will make such a difference over the next few years. Good technology should fit around the broker, not force the broker to fit around the technology.

AI is most valuable across workflows

So far, most practical AI use in broking has happened at the task level. Summarising documents. Extracting information. Drafting an email. Producing notes faster. All of that is useful.

But the bigger opportunity is not one task at a time. It is using AI across whole workflows.

Renewals are a good example. Renewal work is not one action. It is a sequence of actions that includes reviewing history, checking for changes, gathering information, requesting quotes, packaging recommendations and preparing communications. Much of that process is repetitive, rules-based and time-sensitive.

In my view, around 80% of pre-renewal and renewal workflow can be automated or supported by AI.

The important point is not that the broker disappears from the process. It is that the broker’s role changes. Instead of manually running every step, the broker becomes the governor of the process. They review, direct, validate and make the final judgement.

Implementation starts with data and change

This is where many firms still underestimate the work ahead. Thoughtful AI implementation is not just about choosing a tool. It is about preparing the business to use that tool well.

There are two areas that matter most: data integrity and organisational change.

If the underlying data is inconsistent, incomplete or poorly structured, AI will not fix that. It will simply produce poor outcomes faster, often with a level of confidence that can be misleading.

The same applies to migration. Moving off a legacy system is not only a technical project. It is a chance to rethink how work should happen, where human judgement is essential and which steps no longer need to be done manually.

In my experience, firms often underestimate both. Data migration takes longer than expected because no two systems are the same. Change management takes longer than expected because people need time, clarity and support to adopt new ways of working.

Trust, governance and human judgement matter most

The next phase of AI in broking should not be about prioritising handing over decisions. It should be about supporting better human decisions.

That means keeping a human in the loop, particularly for anything customer-facing. AI should not be providing advice. It should not be applying financial judgement on its own. Those responsibilities sit with professionals, and they should stay there.

I come back to a simple principle: trust and verify. Use AI to accelerate work, but always verify the result against the source of truth. If a system cannot show where an answer came from, that should raise concern.

Brokerages also need guardrails around how AI is used, what data it can access, how outputs are reviewed and where accountability sits.

That matters in any industry. It matters even more in one where compliance, evidence and auditability are part of everyday work.

Better advice is the real opportunity

It is tempting to frame AI purely as an efficiency play. It can certainly reduce costs and save time. But the stronger opportunity is to give brokers more time and headspace for the work clients actually value.

That means better conversations, sharper advice, more attention to risk and stronger relationships. Insurance has always been a relationship business. That will not change because AI has arrived.

The industry has largely moved past the question of adoption. If we get that balance right, AI will not replace the broker. It will help create a better one.