The Intermediary – November 2026 - Flipbook - Page 99
T E C H N O L O GY
Opinion
The agent has
entered the building
F
or much of the
past two years, the
conversation about
artificial intelligence
(AI) in banking has
concentrated on what
it might be capable of doing. The
demonstrations have been impressive,
the investment considerable and the
predictions increasingly ambitious.
Yet, for many established financial
institutions, the practical benefits
have remained relatively modest
when measured against the scale and
complexity of their operations.
The emergence of agentic AI moves
the technology beyond answering
questions, generating content or
supporting individual tasks, towards
systems capable of undertaking
activities, coordinating processes and
acting within defined parameters on
behalf of an organisation.
Reconsidering processes
Rather than simply helping complete
an application, an AI agent might
identify missing information, request
documents, validate evidence, initiate
checks and progress an application
through several stages.
The opportunity is to reconsider
how processes operate. For larger
established mortgage lenders, the
difference between a legacy estate and
a modern platform will increasingly
shape what can be achieved.
Many of the largest institutions
have spent decades building systems
to serve millions of customers. Their
technology estates reflect successive
investment, regulatory change and
product development.
Individual legacy systems may
work well, but their data sits in silos,
making connection difficult and
expensive.
Modern platforms are designed
around accessible data and controlled
integration. Older estates accumulate
work around processes designed for a
different era.
Mortgage applications, servicing
requests, product transfers and
variations can involve multiple
systems, repeated checks and
considerable manual intervention.
These activities are necessary, but the
effort involved in coordinating them
can be disproportionate to their value.
Agentic AI can coordinate activities
across a lender’s technology estate,
drawing from different sources and
acting within wider workflows.
That is far easier on an agentic-ready
platform, where data is structured
and available through controlled
connections, than across a fragmented
legacy estate.
Agent team
Agents do not make legacy platforms
irrelevant. Their effectiveness depends
on agentic-ready banking platforms,
well-structured data and controlled
access. This foundation makes it
easier to define authority, understand
activity and validate outcomes.
In legacy environments, every task
becomes harder. Small improvements
in efficiency can produce substantial
benefits at scale.
The opportunity becomes more
interesting when several agents
begin working together. One might
validate documentation, another
check information against policy
requirements and a third coordinate
the next stage of the application.
Together, they manage a sequence that
currently requires intervention across
different teams and systems.
Defining authority
At Ohpen, The ambition is to move
beyond isolated automation towards
intelligent processes that can support
the wider mortgage journey, but
we also recognise that greater agent
autonomy introduces a different
set of questions. Who determines
the boundaries of its authority?
What information can it access,
what decisions can it initiate, and
JERRY MULLE
is UK MD at Ohpen
under what circumstances must it stop
or seek approval? Authority must be
clearly defined, exceptions properly
managed and accountability retained
within the institution.
For larger lenders, these
requirements are particularly
important because even a relatively
small process weakness can become
significant when repeated at scale.
There is, however, an opportunity to
improve control as well as efficiency.
Properly designed agents can operate
within clearly defined rules, maintain
records of their activities and apply
established requirements consistently.
They may also help identify exceptions
earlier and provide greater visibility of
operational processes.
The question is therefore not
whether greater automation
inevitably means greater risk, but
how it can be designed to improve
productivity while strengthening
governance. This is where the next
stage of AI adoption will be decided.
For established lenders, the
opportunity is considerable. Agentic
AI can ease legacy-system complexity
and reduce migration risk by mapping
processes and reconciling data during
change. Yet it is no shortcut around
platform renewal. Success requires
clarity on how agents interact with
banking platforms, where authority
sits and how responsibility is
maintained.
The agent has entered the building.
The challenge now is to give it a useful
job, the authority to perform it, and
the controls that allow the bank to
remain firmly in charge. ●
November 2026 | The Intermediary
97