The industry conversation about AI for financial advisors has a sequencing problem, and it is not a small one.
Most advisors reaching for AI are not chasing technology for its own sake. They want their practice to run more smoothly: client summaries that do not take an hour to prepare, follow-ups that fire without anyone having to remember to send them, and meeting prep that does not eat up an evening. These are reasonable expectations, and AI can deliver on all of them. But it can only deliver if the foundation beneath it holds. Clean, structured data. A CRM that runs the practice. Connected tools that move work without manual handoffs. Build those layers in order, and the efficiency gains are real. Skip them, and you have a capable tool producing faster output from a disorganized base, which still has to be checked, corrected, and cross-referenced against three other systems.
This article is about that foundation and the order in which to build it.
The stack
Think of the operating system of a practice as four layers, built from the bottom up. Each layer enables the one above it. Skipping to the top is how you get a sophisticated tool producing confident output from unreliable inputs.
Layer one: data
This is not a technology problem. It is a discipline problem.
Most advisory practices hold far less structured client data than they believe. Records contain names, contact information, policy numbers, and perhaps a note from the last meeting. What they rarely contain is a complete picture of the household: family composition, dependants, health considerations, business ownership, estate objectives, protection coverage across all carriers, and retirement income from every source.
That information exists in most practices. It lives in email threads, old applications, spreadsheets, and the advisor's memory. It has never been captured in one place in a consistent format.
The starting point is not a new tool. It is a commitment to capture the right information every time and store it in one place. An advisor with 30 structured data points per household in a consistent format is better positioned than one with thousands of records with two fields completed.
The test: if someone opened a client record today, would they know exactly what to do next?
Layer two: the CRM as the spine
The CRM is the system of record for every note, every task, every case stage, and every piece of proof that a step was completed. Not a contact list. The spine.
When a CRM operates this way, it does three things.
Enforces process: Cases do not advance until the conditions for the next stage are met.
Assigns ownership: Every task has a person and a due date.
Escalates issues: When something stalls, the system notices and notifies someone. The advisor does not discover the problem two weeks later.
A concrete example from a recent engagement: An eight-stage new business pipeline for a life insurance practice. Referral received, through discovery, proposal, commitment, application, underwriting, policy delivery, and case close. Every stage has defined entry and exit criteria. Every stage generates the next set of tasks automatically when a case enters it. When a medical result does not arrive within 10 business days, a follow-up fires. When it does not arrive three days after that, it escalates to the next contact in the process. No one decides to chase. The system runs the sequence.
That is the difference between a CRM that an advisor pays for as a phonebook and a CRM that runs the practice.
Layer three: integrations and automation
This layer connects the CRM to the tools around it so that work moves without manual handoffs.
A client receives a booking link. They schedule the meeting that aligns with yours and their calendars, and an auto-confirmation fires. A 24-hour reminder is sent, and the meeting happens. A transcription tool captures the conversation. The summary is pushed to the client’s record in the CRM automatically within seconds of the call ending. No one typed a note twice. No one set a reminder by hand.
A policy reaches its anniversary date. An alert fires in the CRM, a personalized review invitation goes to the client with a booking link embedded. The client books. The advisor sees the meeting in the morning and knows the conversation to prepare for.
A term policy enters the conversion window. An alert fires. An automated renewal email goes to the client. A form is queued for the administrator, and the advisor is notified.
None of this is exotic. The tools required exist in most practices today, many already included in the software being paid for. The gap is not access. It is the willingness to connect the tools in sequence and confirm that the sequence holds.
Layer four: AI
Once the layers beneath it are solid, AI has something real to work with.
Here are three demonstrations from a recent keynote I delivered at the Manulife Wealth President’s Council in Edinburgh, Scotland, on June 11, 2026:
A client analysis took a structured household record and produced a complete advisory summary: Income picture, protection gaps, business exposure, estate considerations, and the three conversations most worth having at the next review. The advisor reads the output, edits where needed, and walks into the meeting prepared in a way that used to take an hour to build.
An operations assistant trained on the firm's own documents: The service guide, workflow checklists, and compliance protocols. Staff ask it questions and receive answers grounded in the firm's actual procedures, not general internet results.
A financial plan generator that took structured inputs and produced a draft plan ready for advisor review. Not a finished plan. A first draft.
Every demonstration ended the same way, with the guardrail stated explicitly. The task the AI performed. The human review step. Where the final output is stored. This is not a formality. It is the rule that makes the tool defensible in a regulated environment where the advisor owns every output that leaves the practice.
Earn the Right to AI Keynote by Emile McLean, Manulife Wealth President’s Council, June 11, 2026
The benefits beyond AI
The result of using AI is well documented: You gain efficiency and more time in your schedule to do other high-value tasks, and your business becomes more profitable. More importantly, getting the build order right produces two additional outcomes that have nothing to do with AI and everything to do with the long-term health of the practice.
The first is audit readiness. A practice where every client interaction is documented in the CRM, where every task has a completion record, and where every stage gate in a workflow has documented entry and exit criteria is a practice that can answer a regulatory inquiry without reconstructing events from memory. The documentation exists because the system required it. Audit readiness becomes a byproduct of operational discipline rather than a separate project run in anticipation of a review.
The second is practice valuation. Advisors approaching succession, a sale, or a book transfer are in a materially different position depending on how systematized their operations are. A practice that runs through the principal's memory and relationships is difficult to transfer and difficult to price. A practice with structured data, documented workflows, and a CRM that runs without the original advisor in the room is a transferable, auditable asset. Buyers and successors pay for systems. They apply a significant discount for practices that exist primarily in someone's head.
The build order that earns the right to use AI is the same build order that protects the practice in a regulatory review and increases its value at exit. These are not separate objectives.
Where to start
Make the CRM the single source of truth for it for thirty days. Pick one repeating workflow, only one. Start small, and do not attempt to automate your entire business immediately.
Once that holds, identify one handoff in that workflow that depends on someone remembering and automate it. Confirm it runs reliably before adding a second.
Once the foundation and the automation are stable, introduce an AI tool to one defined task in that workflow. State the guardrail before using it. Observe the output for two weeks before expanding.
This is slower than buying a tool and hoping it changes behaviour. It is the only version that holds.
This is The Capacity Problem. Practical operations insight for Canadian financial advisors, zero filler. If this was useful, share it with an advisor who needs to hear it.