Financial services firms do not need another adoption percentage to establish that automation is on the agenda. The more useful evidence sits inside the operation: which processes use it, what outcomes changed, where exceptions occur and whether the cost of control leaves a worthwhile return.
The difficult decision is where to automate and where professional judgement remains part of the service.
In financial services, the line can affect service speed, adviser capacity, client trust and regulatory exposure. The effects may not surface together or immediately, so the design needs measures for each outcome that matters.
We're automating where it makes sense. The compliance team reviews everything before we go live. Is that enough?
Compliance sign-off at deployment is one control at one point in time. Design also needs to address edge cases, changing client circumstances, monitoring and the evidence retained for a particular outcome.
Those questions turn a general approval into controls that can operate after launch.
The automation opportunity is real
The administrative burden in financial services can be punishing. I've sat with operations leaders whose teams manually reconcile data across three systems that should have been connected years earlier. One firm we worked with had four people spending most of their time moving numbers between spreadsheets and a CRM that did not integrate. The overhead was real, and so was the opportunity to redesign the work.
High-volume, rule-based work with defined inputs can be a good candidate: parts of KYC or AML operations, regulatory-report compilation, portfolio-data aggregation, onboarding administration, reconciliation, scheduling and reminders. Each still needs assessment against the applicable rules, data, exception routes and consequence of error. A category label does not make the complete process low-risk.
Within those categories, identify steps where inputs are structured, rules are explicit and errors are detectable and recoverable. Establish the baseline and test whether the proposed change actually produces faster processing, fewer errors or lower cost per transaction after the cost of oversight.
Trouble starts when a firm extends the same logic into work with ambiguous rules, unstructured inputs or consequences beyond operations. A control that worked for a bounded screening task is assumed to work for suitability. An onboarding workflow gradually begins to shape advice delivery. Scope expansion needs a fresh decision rather than inheriting the original approval.
The line that matters
At important moments in a financial-services relationship, the client is buying confidence in the adviser and the judgement applied to their circumstances as well as a product or calculation.
I was talking to a wealth management director who put it better than I could. She said: "My best adviser knows that when a client says 'I'm comfortable with risk,' what they actually mean depends entirely on whether their daughter just got married or their business just lost a major contract." A system working from recorded fields may lack that context. The firm should decide where an adviser must surface, interpret and act on it.
Retirement planning for a client whose family situation is genuinely complicated - adult children from two marriages, a business interest, a spouse with different risk appetite. Estate structuring where the emotional dimensions are as significant as the financial ones. A significant investment decision where the client's risk tolerance on paper doesn't match what they're actually feeling in the room. These are moments where the adviser's ability to read the situation - to notice what's not being said, to adjust based on thirty years of context - is the product.
Risk assessment that requires contextual judgement. Regulatory interpretation in novel or ambiguous situations where the automated answer might be technically correct but inadequate for the specific client. Relationship-sustaining interactions where confidence in the adviser - not just the advice - is the value proposition. In financial services, that last category is larger than most operations leaders instinctively think it is.
The integration model that actually works
A useful pattern is to let automation assist preparation while the adviser remains responsible for judgement.
What that looks like in practice: the automated system gathers the client data, runs the compliance checks, generates the draft report, surfaces the relevant information, and flags anything unusual - all before the adviser sits down with the client. So when the meeting starts, the adviser isn't spending twenty minutes reviewing paperwork. They're informed. They're prepared. They can spend the time on the decisions that actually require their expertise.
The intended result is faster, better-prepared service and more adviser time for valuable judgement. Measure whether those effects occur and whether review effort, exceptions or client outcomes reveal a loss of quality.
We helped a wealth management firm structure this kind of integration. Its advisers were spending close to half their client-preparation time pulling together data that already existed in three systems. After automating preparation, one adviser said it felt like getting a day a week back. She used the capacity to go deeper with existing clients, and recorded renewal in her book improved. Those observations need to be read as one account rather than proof that automation caused the renewal change.
The boundary between “prepare” and “decide” must be deliberate, documented and reviewed. Measure the surrounding experience directly: preparation time, missing or stale data, adviser corrections, client effort, complaints, service timeliness and the outcome the process is meant to support. External experience percentages cannot substitute for evidence from the firm's own clients and workflow.
The compliance dimension you can't skip
Financial-services automation must treat compliance as a design constraint tied to the actual activity, firm and customer.
Using automation does not remove the firm's obligations for regulated communications or outcomes. The requirements depend on the activity, firm, customer and delivery model. A system that contributes to suitability or another consequential decision needs controls proportionate to that role and the rules that apply.
The FCA's current approach to AI relies on existing frameworks and highlights Consumer Duty and senior-management accountability. Firms should map those rules to the specific automated process, including how products meet target-customer needs, communications meet information needs and support meets customer needs. Accountability remains with the firm.
Investment management, insurance and corporate finance have different rules and supervisory expectations. Identify the applicable framework rather than assuming a single automation standard covers all financial services.
What this means practically is that compliance review can't be a final gate before go-live. It needs to be embedded in the design from the start. Which automated decisions need human oversight? What are the exception pathways when the system encounters something it wasn't built for? How are you monitoring outcomes on an ongoing basis, not just at deployment? How do you demonstrate to a regulator - six months after launch - that a specific automated decision delivered a good outcome for a specific client?
If you cannot answer those questions, resolve the control gaps before automating the process. An efficiency case does not establish that the firm can govern the resulting outcome.
Unexpected automation can create a trust problem when the client reasonably assumed that a person exercised judgement. Decide what the client needs to understand, what the firm is required to disclose and how the route to human support remains visible.
Telling your clients what you've done
This brings me to something I think too few firms handle well: communication.
Decide proactively what clients should be told about the change and why. A message might explain that portfolio-report preparation has been automated so the adviser can spend more time on interpretation. Test the wording and ensure it describes the real process. Different clients may have different expectations, contractual rights and information needs.
Silence can create credibility risk when a client reasonably expected human involvement. A templated communication or generic response to a complex issue may expose the gap. Give clients a clear route to a person and avoid claiming that automation exists to create adviser capacity if the operating model uses the saving for something else.
That distinction matters enormously.
A word about the workforce
The talent dimension also deserves direct treatment. The standard narrative is that automation “frees advisers for higher-value work.” Sometimes that is exactly what happens, as the adviser example illustrates.
Sometimes the planned outcome is that the same work is done by fewer people. Operations leaders should be explicit about which outcome they are funding and how roles, workload, development and consultation will be handled.
People will compare language about “empowering advisers” with subsequent decisions about roles and headcount. I've seen a firm announce an automation programme in those terms and make redundancies six months later. The sequence damaged internal trust. Communicate the intended workforce effect early, follow the applicable consultation and employment requirements, and update people when the plan changes.
Where to start
If you're reviewing your own automation programme - or about to start one - the question isn't "what can we automate?" It's "what should we automate, given who we are, who our clients are, and what our regulatory obligations require?"
That means mapping every process you're considering against three dimensions: operational efficiency (will this reliably save time and reduce errors?), client experience impact (will this make the client's experience better, worse, or different in a way they'll notice?), and regulatory exposure (does automating this process change where the compliance risk sits, and have we designed for that?).
If you want to work through that mapping properly - and identify which processes need enhanced compliance review before go-live - we've put together a one-page assessment built for financial services contexts. It covers the same categories but adds a fourth column flagging Consumer Duty and sector-specific regulatory considerations for each process. Designed to be something you can share with your compliance officer or managing partner before an automation programme gets approved.
Firms should judge automation by the outcomes and controls of each process, not by how much they have automated. Deliberate boundaries are part of the value case.
