Write an HR consulting proposal by defining the client problem, named deliverables, document and meeting limits, client inputs, acceptance steps, fee basis and exclusions before asking AI to draft wording. Let AI turn those facts into prose, then check every promise yourself. A fixed fee needs a change rule; a retainer needs a monthly capacity and response boundary.
“Review HR policies” is too vague to price or accept. Does it mean reading ten supplied documents, interviewing managers, checking signed files, drafting new clauses or giving legal advice? The proposal should answer those questions in ordinary language. It should also say what happens if the client provides incomplete records or asks for individual case work during a document project.
Collect facts the draft cannot invent
Prepare a one-page discovery record before opening an assistant. Capture the client's size, workforce groups, reason for the project, current documents, approval owner, desired outcome, deadline dependencies and the known risks. Ask what the client expects to receive: a gap matrix, a rewritten handbook, a manager briefing, or ongoing advice. Do not let AI infer that a generic “policy review” includes all of these.
For a 25-person client, the consultant might receive 12 documents, a handbook last approved two years ago and a request to identify outdated instructions. Ask whether those 12 files are the full approved set, whether policies differ by role, and whether the client can supply a version register. A missing approval date is an intake issue, not proof that the policy is invalid. Record it so the proposal can price document sorting if necessary.
Include a data-handling question. A blank template can often be reviewed without names, while a case investigation or signed contract sample contains personal information. Agree who is allowed to provide material, where it will be stored, whether an AI service may receive any of it, and who approves that service. OpenAI and Anthropic both say their business workspaces (ChatGPT Business, Claude for Work) are not used for model training by default, but that settles one question only: client consent, confidentiality obligations and permissions still need separate review.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowState deliverables as files and decisions
A useful scope names the output format, volume and review rounds. “Gap matrix covering six agreed policy areas, with source references and priority notes” is testable. “Improve HR compliance” is not: it promises an outcome the consultant cannot establish from a limited desk review. For a handbook update, say whether the service includes a marked-up copy, a clean copy, an issue log and a staff briefing note. Define who approves the final wording.
Use this blank scope clause as a starting point, replacing every bracketed field before sending. It deliberately separates review from rewrite and avoids implying that AI output is a final professional opinion.
Scope: We will review up to [number] current documents supplied by [date] across [named areas]. We will deliver [file names and formats], each with source references and a list of unresolved questions. The work includes [number] review meeting(s) and [number] consolidated revision round(s). You will nominate [role] to confirm factual accuracy and approve any final policy wording before issue.
The acceptance clause can be equally concrete: “Within [number] working days of delivery, the client will identify factual errors or items that do not match the agreed deliverables in one consolidated response. We will correct those items. New subject areas, additional source documents and changes in policy direction will be estimated separately.” Have a qualified adviser review contractual language where necessary; this sample is an operational drafting aid, not legal advice.
Choose fixed fee or retainer by the work pattern
A fixed fee suits a bounded output: an audit of named files, a handbook revision or a defined workshop document. It needs a document cap, revision cap and assumptions about access to decision-makers. A retainer suits recurring questions and small updates, but only if the client knows the monthly capacity, response window, excluded case types and how unused time is treated. Do not describe an unlimited retainer if one consultant must also serve other clients.
| Fee line | What to specify | Illustrative wording |
|---|---|---|
| Fixed project | Deliverables, inputs, revisions, change rate | One document audit of up to 12 files and one revision round |
| Monthly retainer | Included hours or requests, response target, rollover | Up to six hours of routine policy queries each month |
| Additional work | Rate or separate quote and approval step | No extra work starts until the client accepts a written estimate |
| Expenses or third-party costs | Whether included and how approved | Any agreed external cost requires prior written approval |
Set the fee from your own time estimate, overhead, risk and market positioning. The numbers below are illustrative proposal arithmetic, not a claim about what consultants typically charge. Ask AI to check that multiplication, deliverable count and instalments agree, but do not ask it to decide your price. It cannot know your capacity or the complexity hidden in the client's files.
A filled proposal for a 25-person client
Illustrative worked example: a 25-person bicycle repair business has 12 current HR documents and wants a controlled handbook refresh. The consultant proposes a fixed fee of $2,400 for a six-area document audit, a gap matrix, one marked-up handbook, one clean copy, a 45-minute findings call and one consolidated revision. The consultant estimates 14 hours: two for intake, four for review, four for drafting, two for checks, and two for meeting and revision. $2,400 divided by 14 is about $171 per estimated hour before overhead and risk. That is internal pricing arithmetic, not a contractual hourly charge.
The client will supply the approved documents and nominate one decision-maker within five working days. The consultant will identify unclear or missing source material and may pause the relevant section until it arrives. The proposal excludes individual grievances, disciplinary advice, signed-contract review, payroll calculations, implementation of new software and legal sign-off. A qualified adviser will review any wording where legal interpretation is required. The client approves the final handbook before distribution to staff.
If the client later sends 30 additional personnel files and asks for a case-by-case review, the consultant should not squeeze that into the $2,400 document project. A change note would say: “The new request adds 30 named files and individual case analysis. That work is outside the agreed document audit. We can scope it separately after confirming the data-handling arrangements and the adviser required.” The specific facts are illustrative; the boundary is what makes the fee intelligible.
A separate retainer option might offer six hours of routine policy maintenance a month for an illustrative $900 monthly fee, with work beyond six hours requiring prior written approval. If the client uses four hours, the proposal must say whether two hours expire or carry over. It must also say whether urgent case advice is included. Avoid the vague line “ongoing HR support”, which invites two parties to imagine different services.
Give AI a fact sheet, then challenge the draft
Paste a sanitised fact sheet into an approved workspace. Include only the business type, headcount, document count, deliverables, assumptions and fee terms needed for the proposal. Ask the assistant to write in plain language and list anything it could not determine. Its strongest use is structural: spotting a deliverable without an acceptance test, an exclusion that conflicts with the scope, or a fee line that lacks a payment point.
Draft an HR consulting proposal from the facts below only. Use headings for objective, scope, deliverables, client inputs, fee, exclusions, change control and acceptance. Do not invent legal assurances, client history, deadlines or extra services. Put unresolved points in an "owner to confirm" list. Facts: 25 staff; 12 approved HR documents; audit six named areas; deliver gap matrix and marked-up and clean handbook; one 45-minute findings call; one consolidated revision; illustrative fixed fee $2,400; client provides one decision-maker and current files.An illustrative draft might promise that the handbook will be “fully compliant” and that every manager will be trained. Neither is in the fact sheet. Delete the guarantee, remove the training service and replace it with a limited deliverable description. If the draft inserts a completion date, mark it as a dependency on receiving approved files and client decisions. The finished proposal should reflect agreed facts, not AI's idea of a polished sales document.
Five clause checks that catch costly ambiguity
Document cap: an HVAC contractor says “review our contracts” but has three role templates, two historic versions and 40 signed files. Define whether the cap covers templates or individual records. If you mean only three current templates, say so. The client may reasonably expect signed files to be covered unless you make the distinction explicit.
Revision ownership: a bakery owner sends conflicting changes from two managers after your first draft. A clause requiring one consolidated response from a named approver prevents you from implementing contradictory instructions. Ask the owner to resolve the conflict before the second draft. Count a revision round from the consolidated response, not from each separate email.
Dependency delay: a café cannot find the approved absence policy for three weeks. State what pauses, whether the delivery timetable moves, and which work can continue. A fixed date without a document-receipt condition can make you responsible for a delay you cannot control. Record the missing item in the project log and agree a revised delivery point.
Meeting limit: a client requests six management calls when the fee includes one findings meeting. You can offer a separate estimate or replace some other deliverable by written agreement. Do not assume “reasonable calls” is self-explanatory. Define length, purpose and attendance for the included call.
Use of AI: a proposal that says “AI-powered review” may make the client think a machine will decide legal compliance. Explain that AI may assist with document comparison and drafting, while the consultant verifies source references and the client or qualified adviser approves final wording. State what data categories are excluded from the AI step and how any change to that rule will be approved.
Check exclusions against the sales promise
Write exclusions in plain language, but do not use them to erase the deliverable. If the headline sells “handbook update”, excluding all drafting would be contradictory. Better wording says the consultant drafts revisions to the supplied handbook, while formal legal advice and staff implementation are outside scope. If the client needs those services, name the hand-off. An exclusion should clarify a boundary a reasonable buyer might otherwise misunderstand.
Compare every sentence in the proposal with the scope table. Does “review up to 12 files” match the deliverables? Does the findings call include a written action list? Do fee instalments add to the total? If the illustrative $2,400 project is billed half at acceptance and half on delivery, each instalment is $1,200. Have someone else read the proposal as a buyer and mark any sentence that could mean two things. AI can run this ambiguity check, but the consultant must settle it.
For a source-led audit, use the HR audit checklist to estimate evidence work. The AI policy package tutorial shows a distinct deliverable that can be quoted separately. If the client needs a handbook master rather than one revision, the master handbook tutorial helps define that larger output. For confidentiality questions, the staff-data tool guide provides a decision path.
Send a proposal the client can approve without a call
Include a short assumptions table beside the fee. For the illustrative client, the assumptions might be: the 12 files are in editable or readable formats; one person can answer questions within three working days; no more than six policy areas are included; the client supplies current source wording; and the consultant has permission to use only redacted extracts in an approved assistant. An assumption is not a hidden escape route. If it fails, write down the effect on scope, fee or schedule and ask the client to choose a path before further work.
A blank change-control clause can be copied and completed: “Requests beyond [document cap], [policy areas] or [revision rounds] will be described in a written change note showing added deliverables, fee and timing. Work on the change begins after [named role] approves that note. The original scope continues unless the note expressly replaces it.” This protects the client's ability to decline extra work as well as the consultant's capacity. Keep the note attached to the accepted proposal so the latest agreement is easy to find.
State the client's responsibilities without making the client do the consultant's work. “Provide the current approved handbook and identify a decision-maker” is reasonable. “Warrant that every document is legally correct” would undermine the purpose of a review. Ask the client to confirm which files are live, make staff available for agreed interviews, and decide policy choices within the stated review round. The consultant remains responsible for checking their own draft against supplied evidence.
An illustrative alternative is a four-person accountancy practice asking only for an onboarding checklist. A $2,400 six-area audit would be the wrong shaped offer. A narrower fixed proposal could cover the existing checklist, one updated version, a role-specific welcome-message template and one review call, with no personnel-file sampling. The fee would be recalculated from that smaller scope. The example shows why copying the 25-person proposal verbatim would create needless work and potentially confusing exclusions.
Another client may ask for monthly support but cannot estimate question volume. Offer a trial retainer with a defined cap and a review point. For instance, an illustrative $900 monthly agreement might include six hours of routine document questions, a log of requests and a review after two months. If the first month brings 14 hours of disciplinary case questions, the log shows that the service is mis-scoped. The answer is a separate case-advice arrangement or a revised cap, not an unrecorded decision to answer only some questions.
Check cash flow as well as headline price. A fixed project can have a start payment and a delivery payment, each tied to a stated event. If a client delays approval of the final copy, define whether the second payment depends on delivery of the agreed draft or on acceptance after the review window. That is a contractual choice requiring appropriate advice. AI can flag an undefined trigger, but it should not invent terms or tell you which arrangement is enforceable.
A confidential-information clause should match the real workflow. Say where the client will upload files, who can access them, and what happens to working copies at completion under the agreed retention arrangement. If the proposal says no client data enters AI, do not paste a named contract into a chatbot later. If redacted extracts are permitted, specify who performs and verifies redaction. A proposal that accurately describes the data path is more valuable than a generic assurance of secrecy.
Finally, create a one-page delivery acceptance list from the proposal. For the filled example it might read: six-area matrix supplied, every confirmed finding cites a source, marked-up handbook supplied, clean copy supplied, findings call held, one consolidated revision completed, unresolved legal questions listed. The client can tick each item without treating the consultant's recommendations as a guarantee. If an item is not ready, explain why and record the agreed resolution.
The final page should include the client's decision: accept the named fixed project, request a revised scope, or decline. Add the approval contact, fee, payment milestones, start conditions and expiry of the offer if you use one. Check that attachments are named and that no internal pricing notes or AI prompt history are included. The client should be able to identify the deliverables and boundary without interpreting a marketing phrase.
After acceptance, turn each promised output into a project task and keep a change log. If a new request appears, compare it with the signed scope before starting. The proposal has done its job when the consultant can deliver what was sold, the client can verify it, and neither side needs to guess whether a case, file or extra revision was included.
Further reads
- HR Automation AI Tools: What a Small Business Can Automate Safely — Classify safe automation work before quoting it.
- AI Tools for HR Professionals: Policies, Onboarding and Staff Questions — Compare policy and onboarding services.
- AI Survey Analysis for HR Consultants: Coding Employee Comments Safely — Scope a separate survey-analysis service.
- How HR Consultants Use AI for Policies, Letters and Cases — Separate AI drafting from case advice.
- HR Analytics and AI: Analyse Staff Turnover From a Spreadsheet Export — Add a defined aggregate metrics option.
- Structural Engineer Fee Proposal: Scope and Assumptions Drafted With AI — A worked fee proposal with stages, assumptions, exclusions and variation rates, plus prompts that draft and stress-test each section.
- Consulting Pricing Models After AI: Fixed Fee, Retainer or Value? — Which pricing model survives AI-cut hours? Worked fixed-fee and retainer sums, a value-pricing test, scope limits and wording for telling clients.
- AI Salary Benchmarking for HR Consultants: Sources, Bands and Checks — A repeatable method for turning survey and job-advert data into pay bands with AI, plus the checks that catch invented figures.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: OpenAI Help Center, Data controls in ChatGPT (managed workspaces); Anthropic Privacy Center, Is my data used for model training? (commercial products).