The session combined three complementary perspectives: the president of the Flemish brokers’ federation, a senior executive from a leading insurance brokerage, and a structured presentation from KPMG. Together, they provided a coherent view of where the sector is heading.
1. AI is no longer theoretical: it is already reshaping the broker’s daily reality
A recent market survey shows that roughly half of brokers already use AI tools, mainly conversational assistants, while the other half have not experimented at all. This split is a clear warning: the market is polarizing between early adopters and late movers.
The message was clear: AI is advancing rapidly, and brokers and claims teams can no longer afford to stay passive. Failing to experiment now means falling behind over the next two years.
2. The “4D tasks” show where AI creates immediate value
A useful framework classifies the tasks where AI delivers the quickest impact:
- Dull – repetitive work (document handling, information retrieval, categorization).
- Difficult – cognitive-heavy tasks (offer analysis, multi-company comparisons, fraud detection).
- Dally – slow, tedious tasks (summaries, translations, emails, marketing content).
- Dear – costly tasks (closed-environment chatbots, automated client Q&A, dossier follow-up).
This makes AI tangible even for a small brokerage of 5–10 people: its value is real, immediate, and within reach.
3. Data maturity is the real foundation
AI’s performance depends entirely on data that is:
- Clean;
- structured;
- up to date; and
- consistently captured.
One of the strongest recommendations was to start by fixing data quality. Another key point: “Off the shelf” AI solutions rarely perform optimally without retraining on the broker’s own data, such as contracts, customer segments, historical claims, and communication style.
In short: AI works only as well as your data works.
4. Claims is entering a new era with AI Agents
KPMG presented a shift from isolated task automation to multi-step, autonomous AI Agents that can:
- perceive information;
- understand instructions;
- plan execution;
- take decisions;
- perform actions across systems; and
- self-optimize.
A full motor-claims use case showed three agents working together:
- Intake Agent – collects information, reformulates, completes missing fields.
- Coverage & Liability Agent – checks policy validity, dates, clauses, limits, concludes proportional liability.
- Fraud Agent – detects anomalies and weak signals, compares behavioral patterns, raises alerts.
The claims handler stays fully in the loop, but 70–80% of preparation can be offloaded to intelligent automation.
5. The benefits are massive — even for small organizations
One of the biggest myths in the industry is that digitization only makes sense for large teams. The opposite is now true.
AI allows even small offices to:
- save hours every day;
- accelerate the processing chain;
- reduce inconsistencies;
- improve customer experience; and
- free up people for tasks that require judgement and empathy.
This reshuffles the competitive landscape: AI gives small teams the leverage they never had.
6. The AI Act forces a step change in governance and discipline
The European AI Act introduces:
- extraterritorial scope (even the use of AI output falls under the law);
- four levels of risk; and
- different obligations for those who develop AI and those who deploy it.
Several common insurance use cases fall under or near “high-risk”:
- claims classification;
- email triage;
- pricing in life & health;
- recruitment screening; and
- automated decision support.
High-risk systems require:
- documentation;
- transparency;
- human oversight;
- risk controls;
- audit trails;
- data governance; and
- model robustness.
This is the end of “informal AI experiments”. Governance becomes mandatory.
7. The brokerage perspective: value becomes exponential with volume
The senior executive from a major brokerage firm explained how AI becomes indispensable when dealing with very high volumes:
- hundreds of thousands of calls per year;
- tens of thousands of claims; and
- millions of reimbursements and transactions.
The focus is shifting towards:
- Voice AI to reduce waiting time, understand the reason for calling, and identify the customer instantly;
- smart triage that routes clients to the right expert the first time;
- fully conversational intake (no rigid scripts); and
- multichannel automation across voice, chat, and email.
The message is clear: customer demand is too high for purely human workflows.
8. The real blockers are organizational, not technical
All speakers agreed: the technology is mature. What slows organizations down is:
- lack of aligned leadership;
- unclear or poorly defined use cases;
- insufficient data readiness;
- inadequate governance; and
- fear of starting instead of fear of failing.
The biggest risk today is waiting for a perfect “all-in” AI solution, which could cost 12–18 months of learning time
9. AI as a co-pilot — never as a replacement
One core principle cut across all presentations: AI is there to augment people, not replace them.
AI can:
- synthesize documents;
- draft communications;
- check coverage;
- classify claims;
- detect anomalies;
- monitor workflows; and
- automate repetitive tasks.
But empathy, client relationships, negotiation, judgement, and arbitration remain firmly human. The winning model is “people + AI”, not “AI instead of people”.
Conclusion
The session delivered a strong and united message:
- AI is already embedded in the insurance value chain, even if not always visible.
- AI Agents will reshape claims, brokerage workflows, and customer service in the coming years.
- The AI Act brings structure and discipline, pushing organizations towards robust governance.
- Teams must start experimenting now, even on small use cases.
- And finally: the competitive advantage will go to those who learn the fastest, not those who wait for the perfect timing.
Explore
Connect with us
- Find office locations kpmg.findOfficeLocations
- kpmg.emailUs
- Social media @ KPMG kpmg.socialMedia
