Agentic AI's 2027 budget line belongs behind the advisor

Technology budget season for 2027 has opened, and the three executives who spend their weeks inside the tools advisors will be asked to buy have converged on a message that should shape how the money gets allocated: agentic AI is paying off behind the advisor, in the operational work that surrounds the advice conversation rather than inside it. Joe McQuaid, chief operating officer at Concurrent Investment Advisors, Tom Sullivan, head of product at eMoney, and Conor Curtis, head of product at Practifi, each made a version of that case to InvestmentNews in coverage published September 11, pointing to meeting preparation, client follow-up and the linking work between a CRM, custodial data and planning software as the places the hours come back.

That is a narrower claim than the one this publication has been making about AI's march into the client meeting, where Altruist's Hazel planning agent has put the RIA's core deliverable in line for repricing. The two are not in conflict, and the product leaders' version is the more useful one for planning season: the model's first job inside a wealth firm is getting the file, the follow-up and the handoff ready before the client shows up.

Concurrent's own shape explains why that work is where the pain lives. PWD's records show the firm reported $15.7 billion in regulatory assets across 60 representatives in early September, with 41,105 client accounts spread across 187 employees. That works out to roughly $262 million of assets behind each representative and about 220 accounts per staffer, a ratio at which the binding constraint is rarely judgment about clients—which a good advisor has—but the choreography around every conversation: assembling the household data, drafting the review, routing what comes out of it.

McQuaid's account of the payoff is specific: an agent that reaches into a firm's systems, pulls a household's data, drafts the review and routes the follow-ups, he said, compresses hours of pre- and post-meeting work into minutes. He put prospecting and pipeline work in the same bucket, where research, qualification and personalization happen at a volume a human cannot sustain, and offered the general rule that value appears wherever a repeatable, multi-step workflow is eating advisor and staff time.

Sullivan's contribution is a caution worth posting above the budget spreadsheet: the industry, in his telling, is early in adopting genuinely agentic tools, and a single AI feature is not an agentic strategy, a distinction his team is trying to honor with an always-on planning platform that tracks a client's situation, notices what has changed and raises opportunities when they matter, with judgment and the client relationship still held by the advisor. He framed the ambition as making the system smarter rather than the advisor faster, which sounds like a slogan until you price the alternative.

Curtis described the same shift from another angle: tools that wait to be asked a question give way to tools that raise problems nobody raised. His test of value is the one that should survive contact with a board: not the hours saved, but whether those hours go back into the relationships that produce retention, referrals and growth.

That is where a lot of 2027 budgets will fail, and where the industry's habit of buying automation as an expense rather than a capacity shows up. Hours recovered are an accounting entry until somebody is told what to do with them. This publication has argued that firms buy agent detection without the capacity to act on what it finds and end up paying for the same problem twice; the version now on offer funds the saving and leaves the reassignment unfunded, which is the same mistake made one line item earlier. A firm that frees six hours a week per advisor and lets those six hours dissolve into the same overflowing calendar has bought nothing it can report.

Where the three agree most firmly is on sequence: map the workflow before shopping for a tool. McQuaid's instruction is to find where the handoffs and rework live, to weigh integration as heavily as feature lists because systems that plug into existing data beat ones that force a parallel process, and to judge vendors on the operational layer, access, orchestration and governance—the line, in his telling, between a platform a firm keeps and one it outgrows. Curtis's version is to document how the work actually happens, how an advisor prepares for a meeting and how a follow-up reaches the CRM, before evaluating anything.

Vendors handing that advice to their own buyers is a diagnosis as much as a recommendation: if advisory firms knew the shape of their own processes, the instruction would not need repeating across three separate interviews. The vendors are themselves betting on the workflow rather than the model, as Practifi did when its Sentir release folded AI into the CRM's core and left the open flank to the AI notetakers. Whoever owns the sequence—meeting to note to CRM to planning—owns the renewal, which is why the next recruiting grid in this business runs across the advisor's screen rather than through the custodian's logo.

For Concurrent, the workflow question is a growth question: the firm has built a business converting breakaway recruiting into an equity proposition, backing new teams with capital and infrastructure while handing advisors ownership stakes. Every team that arrives comes with a book that has to be re-papered, re-custodied and re-keyed, and the operational layer McQuaid wants his vendors judged on is the difference between onboarding that scales and onboarding that stalls the advisors already on the payroll.

That makes the scoreboard for 2027 legible: the units are accounts per employee and cost to serve per household, and Concurrent sits near 220 accounts a staffer today. If agentic AI does what the budget line says it will, that number moves within four quarters of the spend and the firm's capacity to absorb the next breakaway team rises with it. If it does not move, the money bought a tidier version of the same week.