The Intermediary – July 2025 - Flipbook - Page 77
T E C H N O L O GY
Opinion
AI’s distant
promise
T
he key challenge facing
technology companies
looking to introduce
artificial intelligence
(AI) has always
been the ‘black box’
characterisation of the technology.
Given the inherently opaque nature
of machine-learning algorithms, isn’t
AI effectively excluded from providing
financial guidance?
The imperative to produce coherent
logic underpinning every step of a
decision has meant that the answer to
this question has historically been a
categorical ‘yes’ – it is excluded. That
is, until now, according to some in the
fintech space. They say that, thanks to
AI’s rapidly evolving capabilities, this
exclusion is fast becoming obsolete.
Apparently, AI has perfected the
collation of exhaustive time-stamped,
tamper-proof and minutely detailed
digital logs. Every customer step
is traceable via inputs, logic, and
outcomes, accountable via tying
decisions to data and goals, and
auditable via time-stamped records.
On request, it is said to be capable
of producing a PDF showing each
step of a client’s journey, with raw
data streams exportable as a CSV. An
application programming interface
(API) for regulators might even allow
decisions to be interrogated – for
example, ‘why was that particular
loan selected?’
Automation to autonomy
Combine this with so-called agentic
AI – where autonomous AI makes
decisions and takes action on behalf of
users – and, the argument goes, you
have a massive game-changer in the
years ahead.
Unlike traditional AI, agentic AI
can set goals, plan, adapt to changing
conditions, and execute complex
tasks independently. It’s an AI with
initiative, capable of reasoning
through problems to produce an
ideal outcome.
In theory, this allows the optimal
selection of a product that arguably
meets all customer care concerns. A
user can ask an agentic AI platform,
for example, to source the best value
£20,000 loan to expand their business.
It would analyse the request to
identify key needs (£20,000, business
expansion, ‘best’ loan), prompt for
preferences (fixed or variable rate,
term), gather data (bank accounts,
credit scores, and business financials,
all with consent), and search eligible
loans, running a cost/benefit
analysis on each.
An optimal loan would be selected,
and the application paperwork
automatically completed, including
all required financials. The
intermediary would then present the
consumer with the suggested loan for
electronic signing.
Promise versus practice
But the key words in all the above are
‘in theory’. Despite the impressive
potential that is outlined, there are
no platforms currently in existence
– or near to existence – that deliver
anything like what I have just
outlined. Some providers might be
utilising elements of AI, but few, if
any, are actually doing so to an extent
that amounts to much.
There is a good reason for this
yawning gap between what could be
and what actually is – to many, the
downside risks still outweigh the
potential benefits.
Technology providers know that
over-dependence on autonomous
systems cuts human oversight,
and that this can lead to biases
embedded in algorithms that
disadvantage or even exclude certain
demographics. Errors or biases in
AI systems might lead to regulatory
breaches or financial harm, raising
questions about accountability and
liability claims.
Even if a qualified adviser remains
acting as an intermediary, sceptics
DAVID WYLIE
is commercial director
at LendingMetrics
Given the inherent
risks [...] the road to AI is
going to remain long and
probably heavy going”
argue that any AI means consumer
agency is subverted, and an AIdriven decision is no guarantee of
‘best outcome’.
On paper, the win for consumers
is optimal product, freed time
and reduced paperwork. For the
industry, it’s more loan approvals,
reduced fraud and default rates.
But, understandably, UK regulators
remain cautious. They describe
themselves as ‘pro-innovation’ and
point to engagement with industry
through the AI Consortium and the
Financial Conduct Authority’s (FCA)
AI Lab and Digital Sandbox, however,
progress has been far from rapid.
There is no AI law currently in place
in the UK, and Labour’s recent move
towards EU re-alignment might mean
us adopting the EU’s restrictive AI Act
at some stage.
Caution ahead
The boom line is that, given the
inherent risks and the absence of a
clear stance from UK Government
and regulators, the road to AI is
going to remain long and probably
heavy going.
The type of all-singing-and-dancing
AI platform that we might think is
just around the corner is still some
years away. ●
July 2025 | The Intermediary
77
