Using AI to Price Agency Work and Write Proposals That Win

Pricing is where many agencies leak the most money without noticing. A retainer set three years ago never gets revisited. A project quote is based on what felt right during the call. A proposal gets copied from the last one, with the client name swapped and a few paragraphs that do not quite fit.

AI will not set your prices for you, and you should be wary of any tool that claims it can. What it does well is help you learn from your own history, spot patterns you would miss, and draft tailored proposals in a fraction of the time. This guide shows how.

The running example is Fieldstone Creative, an invented five-person agency in Boise that builds websites and runs local SEO for law firms and accounting practices.

Start with the data you already have

AI is only as useful as what you feed it. Before touching any tool, gather your last one to two years of proposals and outcomes. For each deal, you want:

• Client industry and size
• Services quoted
• Price quoted and pricing model (project, retainer, hybrid)
• Won or lost, and the stated reason
• Actual hours or effort spent, if won
• Whether the client renewed or expanded

If your proposals and deals already live in a CRM, much of this is one export away. If they are scattered across email and folders, spend an afternoon pulling them into a single spreadsheet. That spreadsheet is your most valuable pricing asset.

Use AI to find patterns in your deal history

With the data in one place, ask AI to analyze it. Useful questions:

1. Which services have the highest win rate, and at what price range?
2. Which industries tend to accept higher prices?
3. Where does actual effort most often exceed what was quoted?
4. What loss reasons come up most, and are they linked to price or to something else?
5. Which clients expanded their scope within a year?

Fieldstone found something it had not expected. Their law firm website projects were consistently won but routinely ran over on hours, mostly because of extra review rounds with partners. The fix was not a lower price elsewhere; it was a higher base price for law firms and a clearer revision limit in the scope.

Treat AI findings as hypotheses. Check them against your own memory of the deals before changing anything.

Sample prompt

Here is a table of our last 60 proposals with industry, services, price, outcome, loss reason, and actual hours for won deals. Identify the three service and industry combinations where we most often underestimated effort. Show the numbers behind each finding.

Move from hours toward value

Many agencies start with hourly math: estimated hours times a rate. That protects cost but ignores what the work is worth to the client. AI can help you think through value, as long as you give it real inputs.

For a client proposal, gather:

• What a new customer is worth to the client, roughly
• How many new customers they need to break even on your fee
• What they spend now on marketing and what it produces
• What the problem costs them if nothing changes

Ask AI to lay these out as a simple value case. A person then decides where to land within a sensible range. This approach tends to support stronger pricing because the conversation shifts from “what does this cost” to “what does this return.”

Offer tiers instead of a single number

A single price invites a yes or no. Three options invite a choice. AI is handy for structuring tiers because it can quickly draft variations of scope.

Tier	Example scope for an accounting firm	Purpose
Essentials	Website refresh, basic local SEO setup	Lowers the barrier to starting
Growth	Everything in Essentials, plus monthly content and review requests	The option you expect most clients to choose
Full service	Everything in Growth, plus paid search and quarterly strategy sessions	Anchors value and suits bigger firms

Keep the tiers genuinely different, not just padded. Each should solve a clearly different level of need.

Draft tailored proposals in minutes, not hours

This is where AI saves the most time. Build a proposal framework once, then let AI fill in the client-specific parts from your discovery notes.

A workflow Fieldstone uses:

1. During discovery, the account lead records the call and takes notes in a set template: goals, current situation, obstacles, budget range, timeline, decision makers.
2. After the call, AI drafts the “your situation” and “our approach” sections from those notes.
3. Pricing and scope come from the tier table, adjusted by a person.
4. A human edits every section for accuracy, tone, and anything the AI invented.
5. The proposal is sent from the same system that holds the deal, so opens and signatures are tracked.

Step four is not optional. AI drafts can include confident claims about results you cannot guarantee. Delete them.

Tools built for quotes and proposals help here, because the finished proposal lives next to the deal, can be signed online, and can trigger an invoice when accepted. Connecting that to online payments shortens the time between yes and first payment.

Follow up with context, not guesswork

Many proposals are lost to silence rather than to a competitor. Set up a follow-up sequence tied to the proposal status:

• Day 2 after sending: A short note asking whether any questions came up.
• Day 5: A relevant example or insight tied to the client’s main goal.
• Day 10: A direct question about timing and whether the scope fits.
• Day 14: A polite close-out that leaves the door open.

AI can draft each message using the client’s own language from discovery. With workflow automation, the sequence stops the moment the proposal is signed.

Review pricing on a schedule

Pricing should not be a one-time project. Every quarter:

• Re-run the win and loss analysis with new deals.
• Compare quoted effort with actual effort for completed work.
• Check whether any tier is rarely chosen and why.
• Review retainers that have not changed in over a year.

Small, regular adjustments are easier on clients than a sudden jump every few years.

Frequently asked questions

Can AI tell me exactly what to charge?

No, and be cautious of tools that promise it. AI can analyze your past deals, highlight where you underprice or over-deliver, and help you build a value case for a client. The final number depends on your costs, positioning, and the client relationship, which is a judgment call that should stay with you.

What data do I need to use AI for agency pricing?

Ideally one to two years of proposals with industry, services, price, pricing model, outcome, loss reasons, and actual effort on won work. Even thirty or forty deals can reveal useful patterns. The more consistently you record this in your CRM going forward, the better your analysis becomes each quarter.

Is it risky to use AI to write client proposals?

The main risk is inaccurate or overstated content. AI can invent results, misread notes, or promise things you do not offer. Use it for first drafts of the situation and approach sections, then have a person check every claim, adjust the tone, and set the scope and price. Never send an unreviewed AI draft.

Should agencies show prices in tiers?

Tiers often work well because they turn a yes-or-no decision into a choice between options. Three tiers is the common pattern: an entry option, a recommended middle option, and a full-service option. Each tier should solve a clearly different level of need, not simply add filler to justify a higher price.

Want proposals, signatures, and payments in one flow? Start a free 14-day trial.