The future of commercial underwriting

The proposal form has had a long run

Commercial underwriting is built on a technology that predates the telephone: the written declaration. A business describes itself, an underwriter prices what is described, and the law holds the description to a standard. The Insurance Act 2015 codifies this as the duty of fair presentation: commercial cover rests on the insured presenting the risk fairly, and a matter is material if it would influence a prudent insurer.

It is worth being fair to the model before examining its limits. Static underwriting was not a mistake. For most of the industry's history, information about a risk was scarce and expensive. A surveyor could visit a site once; nobody could watch it continuously. The proposal form was a rational answer to an information problem, and generations of underwriting discipline were built on making the most of thin data.

But the limits are structural, and every underwriter knows them. A proposal form and a site visit show a controlled, on-paper picture of a risk on its best day. Between renewals the insurer is largely blind. The first signal that a risk was not as described is usually the claim itself, and by then the only question left is how much to pay. In 2024, UK insurers detected 684,800 fraudulent applications at the point of sale and £466m of exaggerated loss at the point of claim (ABI). Those are the detected edges of a wider, mostly honest gap: the distance between what is declared and what actually happens on site.

Risk data stopped being scarce

The assumption underneath static underwriting, that continuous information about a risk is unobtainable, has quietly expired. Personal lines proved it first.

Motor was repriced around observed behaviour: telematics products such as Aviva's Quotemehappy Connect give the safest monitored drivers up to 30% off their renewal premium. In home, Hiscox became the first UK insurer to give every buildings customer a free leak detector, and insurer leak-sensor programmes report around 39% fewer water-damage claims. Flood Re's Build Back Better has insurers fund flood resilience because it can cut future repair costs by as much as 73%. Three different perils, one pattern: when monitoring is cheap relative to claims, insurers fund it, and the data reprices the risk.

Commercial sites are, if anything, richer in latent risk data than homes and cars. A typical industrial site already generates CCTV footage, sensor readings, maintenance records and inspection reports, every day, as a by-product of operating. Almost none of it reaches underwriting. The data is not missing; it is unread. What has changed in the last few years is that software can now read it, live and at scale, starting with the most information-dense source most sites already own: the cameras.

From static assessment to continuous understanding

The first shift, then, is from the snapshot to the moving picture. Continuous monitoring does not replace the underwriter's judgement; it changes what that judgement is applied to. Instead of one controlled presentation per year, the underwriter sees a trend: whether declared activities match observed ones, whether working practices are improving or drifting, whether the conditions that precede losses (unsafe behaviour, blocked exits, water where it should not be) are becoming more or less frequent.

This also changes the meaning of fair presentation. Today the duty is discharged by assertion, checked mainly in hindsight when a claim tests it. Continuous evidence lets a fair presentation be supported rather than merely asserted, which protects the honest majority of insureds as much as it protects the insurer. Disputes shrink when both sides can see the same record.

From fixed price to dynamic pricing

The second shift follows from the first. A price set once a year, on declared information, is the only kind of price static underwriting can produce. Once risk is observed continuously, pricing can begin to follow it: renewal terms that reflect a year of evidence rather than a morning of form-filling, discounts that are earned by demonstrated risk quality rather than negotiated, and terms that respond to sustained improvement.

Motor shows the mature form of this: monitored behaviour feeding renewal pricing as a matter of routine. Commercial lines will not copy it overnight, and there are good reasons (longer policies, more heterogeneous risks, brokered distribution) why the mechanics will differ. But the direction is difficult to argue with, because it favours everyone whose risk is genuinely good. The businesses that run safe, well-managed sites are currently subsidising those that do not, and they are the first to benefit when pricing follows evidence.

From paying for losses to improving the risk

The third shift is the deepest. Consultancies tracking the industry describe it as the move from "repair and replace" to "predict and prevent" (Deloitte): the insurer as an active participant in reducing the risk, not just the payer when it crystallises.

The precedents above already work this way. The leak sensor is not a pricing tool that happens to prevent losses; it is a prevention tool that happens to improve pricing. Continuous monitoring on a commercial site works the same way round: its first job is to warn the people on site in time to act, so the incident never happens. The underwriting data is the by-product. This matters commercially because prevention is the only part of the loop the policyholder feels every day, and it is why funded prevention has been adopted rather than resisted in personal lines: the product got better, not just more observed.

Policyholder expectations will move with this. A commercial customer whose home insurer sends a free leak sensor, and whose car insurer rewards good driving each month, will eventually ask why their business insurer only appears at renewal and at claims. The scale of what there is to prevent is not in question: workplace injury and ill health cost Britain around £22.9bn a year (HSE), and UK property insurers paid out £6.1bn in 2025 (ABI).

What this asks of underwriting

None of this happens by installing software. The harder work is institutional, and it is worth naming honestly:

• Data interpretation becomes an underwriting skill. A trend line of site behaviour is a new class of evidence. Deciding what it is worth, and when it should move price, terms or risk-engineering attention, is judgement work, not automation.
• Consent and governance become product features. Continuous data only flows if the insured agrees, understands and benefits. The ICO's expectations on video and monitoring are demanding, and rightly so. The insurers that succeed will treat privacy discipline as part of the proposition, not a compliance afterthought.
• Product design decides adoption. Personal lines customers accepted black boxes and leak sensors because the product made them worthwhile. Commercial adoption will be won or lost on the same question: what does the insured get, on day one, for sharing their data?
• Brokers gain a role, not lose one. Someone has to advise a business on what data sharing is worth, and negotiate the value exchange. That is advisory work, and it sits naturally with the broker.

An honest five-year view

Predictions in insurance have a poor loss history, so this essay ends with direction rather than dates. Telematics took more than a decade to move motor; commercial lines are more heterogeneous and will move unevenly, with high-hazard sectors (construction, logistics, manufacturing) leading because that is where prevention pays fastest. Some of what is sketched above will arrive slower than its advocates expect, and some of it faster than its sceptics do.

What seems safe to say is this: the information scarcity that made static underwriting rational is gone, the precedents for funded monitoring and evidence-based pricing already exist inside UK insurance, and each year a book runs with continuous data, its owner understands its risks in a way a proposal form cannot reproduce. The direction is visible. The open questions are pace, product design, and who moves first.