Before an award you are committing people you have not yet freed up, to work you have not yet scoped. A model can turn your own historical jobs into a staffing profile in minutes. It cannot tell you whether the person named is competent to do the work, and that is the part the regulator cares about.
Key takeaways
Build the staffing profile from your own closed jobs. Take the last several projects of similar type, extract hours by discipline and by phase, normalise them against a size driver you actually record, and you have a defensible first estimate of who this job needs and when. A model does that extraction and normalisation quickly and shows its working. Then overlay committed hours from live jobs, weight the pipeline by your own historical win rate, and you can see whether the people the proposal names will exist in the month the proposal says they will.
The step that cannot be automated is the competence check. Naming a discipline is arithmetic. Naming a person who is qualified, available and willing to take responsibility for the work is a professional judgment, and it is made by a person.
The Engineers Canada guideline on the code of ethics is a synthesis of the provincial and territorial regulators’ codes. Principle 2 — practise engineering only in areas of their competence, carefully, diligently and with honest conviction — says registrants should only offer services, advise on, or undertake assignments in areas of their competence by virtue of education, experience and ability, and that this includes exercising care and communicating clearly in accepting or interpreting assignments and in setting expected outcomes. It also includes the responsibility to obtain the services of a specialist or expert where required.
Read that against a resourcing plan and the implication is direct. A plan that shows a senior engineer at fifteen per cent for a phase whose real requirement is fifty per cent is not an optimistic schedule; it is a plan to accept an assignment on terms the firm cannot meet. The value of doing the arithmetic before the award is that you find out while you can still change the proposal, decline, or bring in the specialist the guideline contemplates. Your own regulator’s code governs — the provincial and territorial associations each publish their own, and the Engineers Canada guideline is written as general guidance across jurisdictions rather than as the operative rule in any one of them.
Notice that every one of those is a transformation of data you already own. No external benchmark is involved, and none is needed.
When the overlay shows a hole, there are three honest answers: move the schedule, bring in a specialist, or decline. The fourth — name someone and hope — is the one that causes claims. If the answer is a subcontractor or an associate firm, check first that your own agreement with the client permits it: Treadstone Law’s note on subcontracting under an Ontario service agreement sets out when a services contract allows delegation and when it does not, which is not something to discover after the award.
section 74.01 of the Competition Act reaches a representation to the public that is false or misleading in a material respect, made for the purpose of promoting a business interest, by any means whatever. Paragraph (1)(b) separately reaches a statement of performance, efficacy or length of life that is not based on an adequate and proper test — and the section states that the proof of that test lies on the person making the representation.
The practical translation for a proposal: a named team member who will not in fact work on the job, an availability statement the resourcing model contradicts, or a turnaround commitment nobody has tested are all exposures, quite apart from what the client will think. This is exactly where generated proposal copy causes trouble, because a model writes confident capability language by default. Treadstone Law’s overview of misleading advertising rules covers the general shape of the obligation.
An eleven-person Ontario engineering practice is invited to propose on a municipal servicing assignment. The principal’s instinct is that it is “about like the one we did two years ago”.
The firm exports timesheets for the eight most similar closed jobs. A model groups the hours by discipline and phase and normalises them against the one size driver the firm records consistently. Two things emerge that the instinct had missed: the approvals phase on the analogue jobs consumed roughly twice the hours of the design phase in three of eight cases, and every one of those three involved an agency the prospective job also involves. The spread, not the average, is the finding.
Overlaying commitments shows the senior water resources engineer is already committed across the two months the proposal would need her. The firm has three real options and picks two: it proposes a start date one month later than it would have, and it names an associate firm for the geotechnical component, having first confirmed the client’s terms permit it. The proposal does not claim a turnaround the resourcing sheet cannot support. A principal reviews the staffing plan and signs it.
Track the countable things: proposed hours against actual hours by phase on the jobs you win, how often the named team is the delivered team, and how often the approvals phase runs long. Those numbers make the next model better. “Hours saved” does not.
This page is written for a firm that delivers work to a book of clients. If the question is really about the front desk — intake, scheduling, recall and reminders — that lives on the professional practice owners page. If it is about your own month-end, reconciliation and payables rather than what you deliver to clients, that is accounting automation, or bookkeeping automation for coding and receipt capture. The tooling overlaps; the risk almost never does.
Can AI estimate the fee as well as the hours?
It can multiply hours by your rates, which is not the same thing. The fee decision includes risk, strategic value and what you are prepared to walk away from, and none of those is in the data. Keep the hours model and the fee decision separate so you can see which one was wrong afterwards.
What if we have no usable historical data?
Then say so rather than substituting an industry figure. Start recording hours by discipline and phase on the next three jobs and estimate from first principles in the meantime, with the assumptions written down. An estimate with visible assumptions is auditable; an estimate resting on an unsourced benchmark is not.
Does naming a person in a proposal commit us to them?
Commercially it often does, and reputationally it always does. Treat named-team statements as promises and either honour them or explain the substitution before the client notices. The competence expectations in the Engineers Canada guideline on the code of ethics point the same way.
A 30-minute call is enough to tell you whether AI pays for itself here.
