Budget Request Template for an AI Email Tool

A budget request template for an AI email tool has one job: get a yes from someone who will spend ninety seconds on it. Most templates fail because they optimise for completeness. Your approver is not reading a proposal. They are scanning for the number, the risk, and the exit.

This is the one-page version — six blocks, a fill-in prompt for each, and two worked examples at different team sizes. It is what you send when there is no formal procurement process and the decision is one manager's to make.

The short answer: one page, six blocks#

Write six short blocks, in this order, and keep the whole thing under 350 words: the ask, the problem, the cost, the return, the risk, and the review date. Put the ask in the first two lines, because for some approvers that is all they read before deciding whether to read the rest.

The block most requests skip is the risk block, and for an AI email tool it is the one that decides how long the approval takes. A tool that reads a mailbox raises a data question. If you do not answer it, someone else will — three weeks from now, in a thread you are not on.

• The ask — amount, seat count, billing period and start date, in one sentence.
• The problem — the work being done badly now, in hours or in a missed outcome.
• The cost — the annual total, plus anything it replaces or absorbs.
• The return — the break-even threshold, not an ROI multiple.
• The risk — what it connects to, who owns it, and how fast it can be switched off.
• The review date — a named date and the single number you will bring to it.

What your approver is actually deciding#

They are not evaluating the tool. They have no way to evaluate the tool, and they know it. They are deciding whether this is a cheap mistake if it turns out to be a mistake.

That reframes most of the page. A long feature list makes the decision harder, not easier, because every extra capability is another thing they cannot verify. What makes it easier is a small number, a short commitment, a named owner and a clear off-ramp.

This is why reversibility beats return on small software asks. Nobody audits the ROI model on a $900 line item. They do notice a twelve-month contract, an unexplained data flow, or a tool with no obvious way to cancel. Answer those three and you have removed most of the reasons to say no.

The second thing they are deciding is about you. A request with a review date and a number attached says you will report back on whether it worked. That is the part that makes your next request easier, and it costs one sentence.

The six blocks, and what goes in each#

Each row below is a block. Write one to three sentences per block. If a block needs a full paragraph, it belongs in a business case, not in a one-page ask.

State the return as break-even, not as ROI#

Most templates tell you to calculate ROI. Do not. An ROI multiple is a forecast, and a forecast invites an argument you cannot win — your approver has no way to check whether a tool saves 30% of triage time, and neither do you, yet.

Break-even is different. It is arithmetic, and it is checkable. It sounds like this: at our fully loaded rate, this pays for itself if it saves each person 30 minutes a month.

To get the fully loaded rate, take the hourly wage and multiply by roughly 1.43. That ratio comes from the BLS Employer Costs for Employee Compensation release for March 2026: private-industry employers paid an average of $46.60 per hour worked in total compensation, of which $32.60 was wages and salaries and $14.01 was benefits. Cite it in the request — it turns your number from an assumption into a reference, and it takes one line.

Then divide the monthly cost by that rate. The answer is how many hours the tool has to save the whole team each month to be free. Divide by headcount and you have minutes per person. It is almost always a small number, and small numbers get approved.

Worked example: a three-person team#

The numbers below are illustrative. Put your vendor's current published price in before you send it — prices change, and an approver who checks and finds a different number stops trusting the rest of the page.

Worked example: a twelve-person team#

At twelve seats, two things change. You now have a second approver — whoever owns security or IT — and the amount is large enough that a pilot beats a full rollout. Ask for the pilot, and name the expansion trigger on the same page.

Red flags that get a request sent back#

None of these earn a hard no. They earn something worse: a let us discuss, which is how a two-minute decision becomes a six-week one.

• Naming the tool before naming the problem. The tool can wait until the fourth line.
• A monthly number when the commitment is annual. If it bills yearly, lead with the yearly figure.
• No named owner. An unowned tool becomes the approver's problem the day it breaks.
• A trial that already started. It reads as a fait accompli, and for an email tool it means mailbox access was granted before anyone reviewed it.
• Benefits claimed for the team when you are the only user. Ask for one seat instead.
• No off-ramp. State the cancellation terms even when they are generous — especially when they are generous.
• One recipient when there are two approvers. Budget and security are usually different people.

Which tool to name, and when not to name us#

Name one tool and one runner-up. A shortlist of five reads as undecided and comes back with come back when you have picked.

Pick on the criteria your approver can check, not the ones you care about as a user. For an AI email tool that is five things: whether anything sends without a person approving it, whether actions can be undone, whether there is an audit trail of what the agent did, whether the vendor trains on your mail, and how fast the seat count can come back down.

AI Emaily is built around those five, which is why they make a tidy risk block. Copilot mode requires approval before anything sends, agent actions are undoable and written to an audit log, autonomy is a level you set rather than a default you inherit, it connects to Gmail, Outlook and IMAP, and we do not train on user mail — the voice comes from a Personal Context brain you write and per-client profiles you set, not from anything scraped out of your sent folder. Packaging is per-seat monthly with a 7-day free trial on Pro and Autopilot; a card is required and it is $0 if you cancel before day 7. Check the current price on our pricing page the day you send the request. We build AI Emaily, so weigh that paragraph accordingly.

Do not name us in four situations. If your team runs Linux, there is no build — web only. If anyone is on an Intel Mac, the desktop app is Apple Silicon only. If your team works mainly on Android, we ship a PWA there rather than a native app. And if your organisation requires a free pilot before any spend, we cannot be that pilot: there is no permanent free tier, only the 7-day trial. Writing free plan into a budget request that later gets checked is how you spend the credibility you were trying to build.

If the answer is no#

Ask which block was weak. You will usually get one of three answers, and each has a different next move.

If it is timing and the budget is committed for the quarter, ask when the next cycle opens and put the same page in your calendar for that day.

If it is the data question, you did not lose — you found the real approver. Collect the vendor's data-handling page, the list of OAuth scopes the tool requests, and their statement on training, then re-send to both people at once.

If it is the amount, cut the seat count, not the review date. One seat for one person for six weeks is almost always approvable, and it produces the number that makes the twelve-seat version easy.