The Intermediary –- June 2026 - Flipbook - Page 65
T E C H N O L O GY
Opinion
AI, DIY, and the
accountability
question
W
e’ve spoken
with lots of
lenders and
brokers about
AI. Broadly,
people seem to
fall into two camps. The first are still
geing their heads around AI, where
it fits, and whether it maers. The
second are experimenting with DIY
AI tools. They’re building assistants,
connecting workflows, automating
tasks, and seeing exciting results.
Modern AI tools are intoxicating.
In a couple of days, one person with
no coding skills can build something
that would previously have required
several developers and months of
work. That’s a massive shi.
I believe that everyone should now
be doing something with AI, because
the fastest way to understand it is to
start using it. Aer spending over a
year heavily focused on building AI
agents, live in production, handling
thousands of complex lending admin
workflows, here are a few lessons.
1. The easy part is getting
something working once
Most AI demos look impressive. The
harder question is whether the system
still works reliably on the 100th
case, with incomplete information,
changing lender criteria, messy
documents, multiple stakeholders,
and constant back and forth.
In lending, the real operational
drag still sits where it always has:
gathering, checking, fixing, and
chasing information, documents,
and evidence across fragmented
workflows.
AI is exceptionally good at handling
repetitive and consistent operational
work, but moving from a simple
MATTHEW ELLIOTT
is co-founder and chief
commercial officer at Nivo
task to a real end-to-end workflow is
where complexity ramps up quickly,
and closing the gap to go live at scale is
very hard.
2. AI systems do not stand still
One of the most underestimated
parts of AI adoption is how quickly
everything changes. Models improve
constantly. New tools appear weekly.
Existing workflows break. Prompting
approaches evolve. Security standards
move. Integrations need updating.
Edge cases emerge that nobody
predicted at the start. People want to
change and improve things.
What worked brilliantly yesterday
may need redesigning next month.
Staying on top of this is exhausting
and requires concentration, expertise,
resilience, testing, and hours and
hours of work.
Implementing AI is not a one-off
project. It needs continual ownership,
improvement, monitoring, and
adaptation. In many ways, it is more
like managing a team than deploying a
piece of soware.
3. Accountability is needed for
complex technology outcomes
AI is capable at tasks: analysing
information, communicating,
summarising, extracting data,
generating outputs, and progressing
operational work. But there are still
critical roles AI does not replace.
Much of the industry discussion
focuses on human relationships, but I
think there is another role becoming
increasingly important as AI adoption
grows: accountability for outcomes.
Someone still needs to decide:
whether the system is producing the
right results; how changes should be
prioritises; how risks are managed;
how exceptions are handled; and
ultimately, whether the overall
operation is working as intended
As AI systems become more
embedded into lending operations,
this becomes increasingly important.
4. Successful AI adoption
becomes a team sport
In practice, successful AI adoption
usually evolves into a combination
of product thinking, operational
oversight, workflow expertise,
compliance awareness, testing,
monitoring, optimisation, and
security controls.
That does not mean brokers and
lenders should avoid AI. It means the
strategic decision is increasingly about
operating model.
Do you want to build and manage
that capability internally, or work
with specialists who have already
invested in the structures required to
operate these systems successfully at
scale? That is the real DIY question.
AI is coming to lending, and the
operational opportunity is enormous.
The biggest wins still come from the
admin layer: the repetitive gather,
check, and chase work that slows
down applications and consumes
valuable operational time. The
important thing now is not waiting
for perfect certainty. It is starting.
For many lenders and brokers,
partnering with specialists who take
accountability for the AI technology
layer will prove to be the faster, safer,
and more scalable path. ●
June 2026 | The Intermediary
63
