Concierge MVP Guide: Sell the Service Before You Build Software

This guide is for a solo founder, consultant, operator, or technically curious beginner with an idea for software, AI automation, or a productized workflow. You suspect a business would pay for the result, but you do not yet know the exact inputs, edge cases, approval steps, or price.

The first customer is not everybody who uses software. Pick a narrow operator such as a small web agency owner, independent property manager, bookkeeping firm administrator, or local trade business. Then choose one repeated outcome they already need, such as a Friday client update, a cleaned quote-request summary, a listing information check, or an owner-approved report draft.

Avoid workflows involving medical decisions, legal advice, employment decisions, financial recommendations, safety-critical actions, or sensitive customer communications unless qualified human review and appropriate controls are already in place. A beginner needs a low-risk workflow where the buyer can approve the final output.

A concierge MVP delivers the proposed value manually and openly. The buyer knows a person is doing the work. You might use a form, spreadsheet, document, email, and a few AI-assisted drafts behind the scenes, but the customer is buying an outcome rather than a finished app.

This is different from secretly making a service look automated. Do not invent a dashboard, claim instant processing, or imply that an autonomous system is handling work when you are doing it manually. Transparency gives you room to ask questions, learn the messy workflow, and keep approvals where they belong.

The useful evidence is paid use, completed delivery, repeat demand, buyer behaviour, corrections, and a clearer process. Compliments, wait-list emails, and friendly survey answers can help, but they do not prove that the workflow is valuable enough to buy.

Start with a workflow the buyer already performs badly, slowly, or inconsistently. Look for repeated input collection, sorting, checking, summarising, drafting, routing, or reporting. The output should be easy to inspect and useful even when a human produces it.

Write the offer as: I help [specific buyer] turn [messy repeated input] into [approved output] every [cadence]. For example: I help small web agencies turn scattered project notes into an owner-approved Friday client update for up to five active projects.

Keep the outcome narrow. A promise to automate agency operations is fog. A weekly draft covering progress, blockers, decisions, and next steps is a deliverable. The buyer can understand it, inspect it, approve it, and tell you what is wrong.

A useful first pilot usually lasts one or two weeks and has hard volume limits. State the number of outputs, input sources, projects or records, delivery days, turnaround, approval rounds, and exclusions. Decide what happens when the buyer sends late or incomplete information.

A two-week agency update pilot might include four draft updates across two Fridays, up to five source notes per project, one owner approval round, and an exception log. It should exclude direct client sending, project management changes, meeting attendance, unlimited revisions, new software setup, and access to unrelated inboxes or files.

Put the scope in a short written proposal before delivery. If the first week reveals a different workflow, change the next cycle deliberately. Do not absorb a new department, extra channel, or new customer-facing action because the buyer says it will only take a minute.

A practical AUD starting range for a capped two-week manual pilot is $250-$750. Use the lower end for a simple output with clean buyer-supplied inputs and little access. Move higher when the workflow has more records, messy source material, several stakeholders, tighter turnaround, or more checking and approval work.

A managed weekly workflow can start around $500-$2,000 per month once you know the real delivery time and volume. A separate automation mapping sprint can start around $500-$1,500 when the buyer wants a documented process, exception list, risk notes, and a prioritised backlog after the manual pilot. These are quote-building examples, not market rules or income claims.

Calculate a private floor before quoting. Estimate intake, delivery, QA, communication, revisions, reporting, admin, and a buffer for the first unfamiliar cycle. Add direct tool costs. If your fixed price does not cover the realistic work, narrow the pilot or raise the fee. Do not call unpaid labour validation when the buyer receives a useful business output.

You usually need less technology than you think: a simple intake form, spreadsheet or tracker, document template, calendar, cloud folder, and email. Add an AI tool only for a bounded draft or classification step that you can inspect. The workflow should still make sense if the AI tool is unavailable for a day.

Collect the minimum data needed to deliver the pilot. Ask the buyer to provide selected notes or exports instead of granting broad inbox, drive, CRM, or customer-account access. Use separate client folders, strong authentication, and a written deletion or handoff step. Never paste confidential material into a tool unless the buyer has approved that use and the tool is suitable for the data.

Keep irreversible actions manual. Do not auto-send customer messages, change records, approve refunds, publish content, or trigger payments during an early experiment. Draft, flag, and recommend. Let the authorised buyer approve the action.

List 20 operators in one niche. Look for visible signs of repeated admin: many active client projects, frequent listings, a busy quote form, a small team, regular reports, or public comments about slow updates. Use published business contact details and research each prospect before writing.

Lead with the outcome and a question: Hi [Name], I am testing a small manual service for web agencies that turns scattered project notes into owner-approved Friday client updates. It is a two-week capped pilot, not new software. How are you handling weekly client updates across active projects now?

If the problem is real, offer a short call to map one current cycle. Do not promise hours saved, fewer errors, happier clients, or a future app before the pilot measures anything. Sell the small paid delivery and agree on what would make the test worth continuing.

For each cycle, record the input received, missing information, manual decisions, clarifying questions, draft time, QA time, approval changes, final outcome, and any step that repeated. This is the evidence that turns a loose service into a reliable workflow.

Separate standard steps from exceptions. Standard steps happen in nearly the same order each time. Exceptions need judgement, missing context, permission, or escalation. Automating a stable formatting step may be sensible. Automating a judgement-heavy exception because it is annoying is how bad systems get built.

Send the buyer a short end-of-pilot summary: what was delivered, what changed after approval, how long each cycle took, where inputs broke, what should stay manual, what could be templated, and whether the next offer is another managed cycle, a narrower pilot, or no further work.

Do not build because the manual work felt tedious. Build when the same valuable step repeats, the input is predictable enough, the output can be checked, and the buyer wants the outcome again. One paid pilot is evidence for the next test, not proof of a large market.

A simple decision is to keep delivering manually, narrow the buyer, change the offer, stop the idea, or automate one low-risk step. For the agency update example, you might template the approved update structure or automate collection reminders while leaving judgement, missing-context questions, and client sending with the owner.

The goal is not to build a startup on schedule. The goal is to get a real buyer, learn the real workflow, and avoid spending months polishing a solution for an imaginary process. Want the outreach script, pilot template, intake form, validation scorecard, prompts, and automation backlog? Open the Manual Workflow Concierge launch pack and customise the working files before contacting buyers.