Put five candidate workflows on a table. First, the receipt, filing and record-keeping that follows every signed engagement letter and return. Next, the annual company-statement cycle, where a regulator invoice is matched to the client, paid or passed through and filed before a deadline that carries a fine. BAS holds the largest block of recurring hours in the practice across reconciliation, review and lodgement. The quarterly CFO meeting requires booking, preparation and the conversion of notes into actions. Client onboarding has the most variation because every client arrives differently.
Every firm weighing AI seriously ends up with a version of this table. Most of the candidates can be partly automated. Management still has to decide which moves first, which follows later and which should remain unchanged for now. A roadmap supplies the evidence for that ordering.
What the evidence says about each
Look closely at the five and they stop being interchangeable “automation opportunities”. They differ on exactly the dimensions that should drive the order.
The signed-document tail is small, bounded and frequent. Its steps identify the item, file and name it, and update the record. Anything unmatched follows a clear exception path. The component can be tested quickly and gives the firm evidence about the operating pattern that will follow.
The company-statement cycle is the cleanest rules case of all, and its value is not mostly hours. It is consequence avoidance: a missed deadline has a price printed on it. Work like this earns its place on a roadmap even when the time saved looks modest.
BAS carries the largest block of recurring hours among the five, which creates a strong capacity case. It also has dependencies and non-negotiable human gates. The work can start only when the reconciliation is ready, while peer review and lodgement stay with people. Systemising BAS therefore requires a staged project.
The quarterly CFO meeting stays entirely human. Booking, preparation and the conversion of notes into recorded actions can be systemised around it. The main value is continuity in the client relationship rather than hours returned.
Onboarding contains the greatest variation and judgement. It remains a genuine opportunity, recorded for later review, but it is a poor first move for proving the operating pattern.
What a roadmap has to capture to support that reading
None of the reasoning above is available from a list of workflow names. It comes from evidence recorded for each workflow: frequency and volume, active effort and elapsed time, the impact of improvement, current readiness and the step detail underneath. Readiness covers whether the inputs, systems and process can support systemisation now.
The step detail matters more than any other layer because the automatable share lives there. BAS shows why. At workflow level it looks like one large candidate. At step level it separates into a readiness check and document preparation that rules can run, followed by peer review and lodgement actions that stay with people. A step map shows both the systemisable share and the human gates. A list of workflow names cannot provide that evidence.
Our AI Roadmapper captures this information with the people who run each workflow during an AI Strategy and Roadmap engagement. The resulting evidence maps potential impact against the effort the firm can carry and shows which candidates require substantive build work. The signed-document tail is likely to appear early, BAS as a staged sequence and onboarding later with the reason recorded.
Evidence and judgement, both on the record
The order cannot be calculated from the capture instrument alone. Evidence narrows the field and the recommendation logic proposes a sequence. An accountable person then decides because engagement shape, team appetite, seasonal pressure and a partner’s judgement about a client relationship also belong in the decision.
Roadmapper generates a deterministic recommendation for each shortlisted workflow from the captured evidence. The consultant confirms or overrides it and records the rationale either way. The roadmap therefore shows what the evidence suggested, what management decided and why. Someone revisiting the plan in six months can read the reasoning rather than reconstruct it from memory.
The roadmap is a point-in-time plan: the best-evidenced sequence available on the day it is produced. It should be revisited as work completes and the facts change. Detailed costing of build-and-run economics follows on the shortlist, where it can be done properly instead of superficially across every candidate.
The order also has to survive the calendar
The sequence also has to fit the firm’s planning horizon. If the first two stages need eighteen months of attention and the firm budgets and commits in twelve, management has a mismatch to resolve before the build starts.
The horizon is captured at the start of the engagement and tested against the proposed stages. If the immediate stage fits and the next does not, the firm can narrow the stage, extend the horizon or accept the stretch with the consequence recorded. Resolving it then is cheaper than discovering it halfway through a build.
A deferral also needs a reason. Onboarding might wait because variation is high and the operating pattern remains unproven. Recording that reason tells the firm what evidence would justify bringing the work forward.
A sequence of decisions, not a statement of intent
The roadmap should leave the firm with an order it can act on. Start the signed-document tail because it is bounded and provable. Include the company-statement cycle because the rules are clean and the consequence of delay is known. Stage BAS with its gates intact. Hold onboarding until the operating pattern is proven.
A roadmap becomes useful when it resolves the order of work and records why each decision was made. Evidence supports that order; accountable judgement makes the final call.
If your firm has the intent but not yet the sequence, get in touch.
Damien Zeller is the Principal of VonBolt Technology Group.
