How funders should prepare for the AI revolution

Artificial intelligence is transforming the way brokers and customers approach the market, and radically automating processes

Artificial intelligence threatens to disrupt the long-standing hierarchy of lenders and render some high profile banks and asset finance companies invisible to customers.

The algorithms and machine learning technology used by these sophisticated engines give customers useful new tools to conduct a far more granular search for the right funder both for their profile and the asset class they are looking to finance, but could also completely overlook long-established lenders.

The challenge for every lender is to ensure their offerings still appear on the radar of customers and brokers when algorithms patrol the online world looking for answers.

“Within 24 months, I think it’s viable that most SMEs, and definitely the intermediaries that we work with, will be assisted, at least in part, by AI agents, like Copilot or ChatGPT,” said Powell.

These agents are software programmes that use artificial intelligence to carry out tasks autonomously. They act on the goals set by customers and use large language models (LLMs) to understand the task, to analyse it, and then to present an outcome.

A typical case might see a customer ask an AI agent to find the best monthly cost or lowest total cost of finance for a specific Ford Transit van.

Taking this one step further, the AI agent could contact a broker or lender directly, eliminating the initial human contact. As the LLMs absorb more data, they could even start to assess the eligibility criteria of different funders, said Powell.

This means lenders with unclear or opaque criteria might not appear at all in these search results.

“Funders in the direct space that might have previously invested in getting their name further up Google search results in SEO are going to have to do the same for generative engine optimisation,” said Powell. “So review your offering. Is it clear? Is it available? Is it out there? Who do you want to be reaching? What are the routes that you would like people to be coming to you for deals? You’re going to need to look at your product offerings and make sure that what you have published is not only readable and very clear, but is also up to date.”

This brave new world risks upending decades of relationships in which brokers have learnt the different appetites of different funders for different deals, even if none of this is officially documented.

One solution is “to think like a machine, with strong questions and answers,” said Andy Taylor, Sales Director, Haydock Finance.

“It’s more of a dark art than SEO [search engine optimisation],” he added. “You have to put in the hard yards to provide original, strong content on your website, explaining why you’re an expert in what you do.”

While investing in this approach will keep funders in the conversation, arguably the bigger return on investment for lenders is the transformation in internal efficiency promised by AI, starting with onboarding.

“If you’re a funder and you’ve invested in your digital technology stack, you’ve got clearly defined requirements for what you expect from an application,” said Powell. “An AI agent is going to be able to collate for the customer all the information required for that deal when it’s proposed. That changes the expectations of right first time.”

Investing in broker relationships to make them digital partners is a key piece of this jigsaw and offers the best “bang for buck,” she added.

Only about one in 10 sets of documentation received by Haydock are “right first time and ready to pay,” said Taylor, but those that are could fly through automated decisioning, leading to slicker, quicker payouts.

While this might seem to threaten both customer-facing and back office jobs, Taylor emphasised how the technology holds the potential to free staff from mundane, repetitive tasks so they can focus on deals and contracts that require greater levels of human involvement.

“We thrive on interactions with people, and that will be here forever, but where this can definitely help is [identifying] the 20% of definite ‘yeses’ and the 20% of definite ‘noes’. And if you can get AI to deal with those, then it gives you more time to look at the more difficult cases, the ‘grey’ cases,” he said.

But with lenders under pressure from customers and brokers to keep pace with their AI tools, cyber and security risks increase, cautioned Antony Clegg, SVP, Product Management, Odessa.

“How do you wire together these different MCP servers [Model Context Protocols are programs that connect AI models to external data and tools], with the AI models that the broker or the end customer are using in such a way that it doesn’t open security holes? There are unique security challenges,” he said.

Criminals are already using AI to target IDB tools with scams, and Richard Huston, Managing Director, VAMOS, warned lenders to be very cautious about the trust they place in AI agents. The systems are trained to ‘believe’ everything they read, a fundamental weakness that fraudsters will exploit.

“There’s a huge risk both in terms of defrauding customers, and of fraudsters targeting lenders as well,”he said.

Meta’s rule of two offers a potential safeguard.

“If you have agents which have insecure inputs into their prompts, access to sensitive data and the ability to transact, then you’re in a very dangerous situation,” said Hughes. “So the idea is that you have at most two of those three.”

Alternatively, “just test the hell out of it. There are some very good test frameworks out there to test this sort of stuff,” he said.