Pierre du Toit (pictured), chief data officer at ManyPets, did not arrive in insurance through the usual commercial route. A mathematician by training and a musician by inclination, he has spent his career building analytics, data science and AI capabilities across insurance businesses. At ManyPets, that work is now reshaping claims, pricing, servicing and distribution, with pet insurance showing what an AI-enabled insurer looks like.
Originally from South Africa, Pierre du Toit studied mathematics and actuarial science at university, alongside music, before beginning his career in financial services – first in banking and later moving into insurance after relocating to the UK.
He joined Vitality, where he spent around a decade in a range of roles, eventually becoming Chief Analytics Officer. During that time, he built the company’s data science function and was working on AI models well before they became widely adopted across the industry.
As du Toit puts it, “I studied it because I liked mathematics,” and at the start of his career “it was simply called statistics, taught by some incredibly boring professors at university.” What has changed since then is not his core interest, but the commercial potential of the discipline. Over time, he says, data and AI have become a field “where, with data and AI, you can create fantastic products and services and add a lot of value.”
He joined ManyPets two years ago, at what he describes as “a really interesting time” for the business: a period of scaling, with enough data accumulated to begin using it in more sophisticated ways.
At ManyPets, du Toit says the company does not frame its agenda as a generic digital exercise. “We don’t really talk about ‘going digital’ in isolation,” he says. “We focus on how AI and data improve the core of how the business operates.”
In practice, that means focusing on the moments the customers notice most - claims, support and response times. The aim is to reduce friction, remove unnecessary hand-offs and make processes clearer, while preserving human involvement where cases are complex or sensitive.
For customers, he says, the outcome is “better, more friction‑free journeys at the moments that matter most”, particularly when making a claim or asking for support - delivering faster, higher-quality experiences around the clock.
The clearest example of this thinking is claims processing. ManyPets handles hundreds of thousands of claims a year, and du Toit says the process used to be heavily manual. The company’s answer has been an AI claims assistant that screens every claim it receives.
Internally, the group of models behind that system is known as Millie. The models were trained on historical claims data to determine what action should be taken when a new claim arrives, whether pre-existing conditions are present, whether deductions are needed, or whether the claim can be approved as submitted.
According to du Toit, the AI assistant now screens 100% of claims, with 55% paid on the same day. Around 80% of claims are paid within five days, up from roughly half that level less than a year earlier.
For a pet insurer, speed is not just a productivity metric. Claims often arrive at moments of stress, when an owner is dealing with a sick animal and a veterinary bill at the same time. Du Toit argues that faster settlement has therefore improved not just efficiency, but reassurance and trust. Customer feedback, including Net Promoter Score and Trustpilot responses, suggests that the change has been noticed.
Claims are only one side of the equation. Du Toit says another major area of investment has been pricing, specifically around getting costs and loss ratios under tighter control.
Over several years, ManyPets has built up a proprietary dataset and rich feature set across customers, pets, health and wellness, giving it a uniquely detailed understanding of pet risk. This foundation supports training for more advanced models, enabling more accurate risk assessment and fairer pricing.
The commercial outcome matters as much as the technical one. Du Toit says the pricing work has supported significant year-on-year new business growth and has also helped turn the company profitable. In that sense, AI at ManyPets is not being positioned merely as a service enhancement tool, it is also being used to sharpen underwriting economics.
The insurer has also extended AI into other parts of the business. In marketing, du Toit says the company is using it to improve and standardise tone of voice, test new propositions with synthetic audience segments, and support campaign development.
He points to the company’s “TailMates” campaign as an example of AI being used in a more creative setting, generating video “tails” for pets and their owners. It is a less operational application than claims automation, but one that shows how broadly the company now sees the technology.
Elsewhere, he says AI is already embedded in engineering, analytics, customer servicing and reporting. Engineers and data scientists use AI agents to accelerate coding. Customer service teams use it to route inbound messages and support faster responses. In reporting, it is used to flag anomalies and improve access to insight for non-technical users.
His summary is blunt: “we’ve embedded AI across every part of the business”.
Du Toit’s view of technology strategy is also clear. ManyPets, he says, starts with customer value rather than supplier relationships. The first question is what customers actually need, and only then does the business decide whether to build or partner.
Its natural instinct is to keep core capability in-house. “We’re fundamentally a tech company and we have a very strong internal engineering team,” he says. By building rather than outsourcing, the firm keeps control of its data and technology stack and makes the business more sustainable over time.
That does not mean refusing external support where it makes sense. Du Toit identifies AWS as a key hosting partner and Google as the provider behind the company’s data platform capability. Those are foundational services ManyPets does not see as worth replicating internally, but the core systems that matter competitively, including policy administration, claims and AI model training and deployment, are more tightly controlled.
One of du Toit’s more pointed observations concerns implementation risk. He argues that many firms misunderstand what AI can do when underlying workflows are weak.
“One of the first things we realised is that AI does not fix a broken process,” he says. “If you bolt AI onto a badly designed workflow, you can actually make things considerably worse.”
That has led ManyPets to focus first on workflow redesign, sometimes in relatively modest ways, before introducing automation. In his account, the gains have often come not from grand reinvention but from cleaning up the process and then applying AI to it intelligently.
He also says resistance inside the business faded quickly once teams saw that AI could remove repetitive and unpopular tasks. Adoption, he suggests, accelerated not because of abstract enthusiasm for the technology, but because it proved useful.
Rather than undertaking sweeping restructures, ManyPets has used cross-functional teams that bring together product, data and engineering. The model is built around clear ownership and clear outcomes, with teams formed to deliver against specific business goals.
On training, du Toit says the company has avoided a generic one-size-fits-all approach. The way a back-end engineer uses generative AI is different from the way a product manager, marketer, finance analyst or data scientist uses it. As a result, training has been tailored to the workflow and role in question.
That pragmatic approach extends to the broader architecture. Du Toit says good data has been essential, but so too has the ability to move models into production. ManyPets has invested in the infrastructure needed to make data accessible and reliable, then linked that to deployment capability and core systems via APIs so that model outputs can flow directly into operational processes.
The result, he argues, is an end-to-end chain from data to model to business outcome.
Asked where AI has made the greatest difference, du Toit does not isolate a single function. Instead, he points to gains across the whole enterprise.
In engineering and analytics, coding productivity has improved sharply, with work that once took weeks now often completed in an hour. In claims, AI extracts information from veterinary notes and invoices and either automates decisions or structures cases for handlers more efficiently. In pricing, it improves risk prediction. In marketing, it supports content generation, campaign execution and targeting. In customer service, it assists with routing and response. In reporting, it helps identify data quality issues and gives business users faster access to insight.
For insurance professionals, the broader point is that ManyPets appears to be treating AI less as a standalone function than as an embedded operating capability.
If there is one management lesson du Toit draws from the past two years, it is the importance of focus. He says the business has no shortage of ideas, but too many potential uses for data and AI can become a problem in itself.
The challenge, in his words, is “saying ‘no’ to even good ideas”, choosing a small number of priorities and executing them well. He also says the projects that performed best were the ones anchored from the start around a clearly defined metric. Where teams aligned around one measurable outcome, delivery stayed tighter and impact was stronger. Where that discipline was absent, progress was slower and results weaker.
For insurers wrestling with sprawling transformation portfolios, that may be one of the more familiar observations.
Looking ahead three to five years, du Toit sees several developments shaping insurance.
The first is the rise of personal AI assistants that understand not just general information, but the individual user: their documents, preferences, emails and working style. The second is what he describes as “auto‑research” or automated model development, where AI systems iterate through modelling ideas continuously and discard what does not work. The third is richer, real-time insight for business users, supported by smarter models and faster decision support.
He also expects large language models to become an acquisition channel in their own right. Today, he notes, general insurance distribution remains heavily reliant on aggregators and direct marketing through Google. In time, he believes models such as ChatGPT, Claude and Gemini could become meaningful routes into insurance purchasing.
In pet insurance specifically, he expects a more connected model to emerge, linking insurance, advice, triage, claims and treatment more tightly into a single customer journey.
Du Toit is notably relaxed about the direction of travel following the Competition and Markets Authority review into veterinary pricing. He says ManyPets welcomes moves that make pricing clearer for consumers and more structured for insurers.
At present, the firm receives invoices from vets and extracts cost information from them, which he describes as a difficult process when trying to benchmark and manage cost effectively. If greater transparency leads to cleaner, more structured data, he believes it will strengthen existing capabilities rather than disrupt them.
For ManyPets, du Toit says, the regulatory direction is not a disruption but an opportunity. Improved data quality would strengthen existing pricing and claims capabilities, rather than require fundamental change.
Away from insurance, du Toit returns to music. He plays both classical and electric cello, usually late at night, and describes music as having been a major part of his life.
He is also using AI in that world, applying it to the creative process to generate ideas and material for compositions. That, he says, feels less like a departure from his day job than an extension of it. The same mathematical foundations are there; only the application changes.
Family life, however, has imposed its own practical limit on extracurricular ambition. With a three-year-old daughter at home, he says free time has become “a bit of a luxury”, though clearly not enough of one to keep him away from either insurance or the cello.
