AI Business Ideas

3 GitHub Repos for Selling $999 AI Assessments

Which problem would a small business pay you to investigate this month? In his video on three GitHub repos, Corey Ganim proposes a $999 diagnostic for each of three problems: repetitive work eats the week, AI search answers overlook a brand, or website inquiries get lost. The repositories help collect evidence. The paid work is deciding what that evidence means and what the owner should fix first.

The offer in one sentence

Pick one costly workflow, agree on how you'll measure it, and deliver a short report with one recommended fix and a way to verify it. Keep implementation outside the assessment. Start before 2027 with one buyer and one measurable problem, not a promise of revenue by a deadline.

Watch Corey Ganim's Breakdown

Three Repos, Three Different Buyers

RepositoryWho buys and what they getWhat could come next
ActivityWatchAn operations lead gets a ranked list of avoidable manual work, validated with the people doing it.Redesign one workflow and remeasure time spent.
ElmoA local business or marketing lead gets a sampled comparison of AI mentions, citations and factual errors.Repeat a fixed set of tests and report meaningful changes.
Browser UseA business paying for inbound leads gets a timeline showing where approved test inquiries slow down or disappear.Fix routing or alerts, then check that the fix holds.

ActivityWatch: Find the Work That Eats the Week

At 01:01, Ganim suggests an assessment for a small team that feels busy but can't say which tasks consume the most time. ActivityWatch is a local-first time tracker. Despite the video's shorthand about recording the screen, the standard watchers log metadata such as the active app, window title, browser tab title and URL, and whether the user is away. That's not screen video, but a title or URL can still reveal client names or sensitive work. Agree on what is collected and retained before a voluntary pilot. Keep raw logs local and use approved, redacted totals in the report.

Start with the owner and a few representative workdays. Ask them to label the workflow: an hour in a browser might be invoice reconciliation, customer support, or research. Pair the totals with a walkthrough, then rank up to five confirmed bottlenecks by time spent, likelihood of improvement and effort to fix. Report measured baseline and estimated savings separately. If a shared template would do the job, recommend it before an AI tool.

Elmo: Test Whether AI Answers Mention the Client

At 05:01, Ganim proposes testing how a business appears against three direct competitors. Elmo tracks mentions and citations for a defined set of prompts and supported engines. Agree on a service, location and buyer intent. Run the same small set of questions more than once, log the engine and date, and save representative responses and cited URLs. In the report, show a denominator: for example, four mentions in 20 sampled answers, rather than a claim that the client is visible to 20% of buyers. Check whether cited pages actually support each claim.

Self-hosting still takes infrastructure and can incur provider costs. As of 18 November 2026, Elmo's $29/month Cloud Starter lists one brand and ChatGPT-only tracking. Confirm coverage and limits before quoting a multi-engine service. Ganim's $300/month monitoring example makes sense only if the report helps someone decide what to correct or update. A sample of model answers cannot prove traffic, attribution or future rankings.

Browser Use: Check What Happens After Someone Asks for a Quote

At 07:43, Ganim proposes a seven-day inquiry audit. Browser Use is an MIT-licensed Python library that lets an agent operate a browser; hosted browser and agent usage are separate. Start manually unless automating repeat runs would actually save time.

Get the client's written approval for specific pages, times and test accounts. Use staging or a client-approved test route and synthetic details. Check when a test inquiry is acknowledged, when it appears in the internal queue, and when someone sends a useful reply. Keep automated checks read-only where possible, and have a person handle any approved submission that triggers messages, bookings or CRM updates. Record the count of test inquiries, failures and delays. That shows how the sampled paths behaved, not how many real leads were lost. To estimate real impact, validate the pattern against client-approved aggregate CRM data.

The Simplest First Assessment Doesn't Need a Repo

At 11:41, Ganim describes a 45-minute owner interview followed by three to seven off-the-shelf suggestions. For your first project, narrow that to one recurring workflow. Document what starts it, what people actually do, and where the handoff stalls. Deliver a one-page recommendation with its assumptions, one success metric and a review call. Use redacted notes; get permission before recording or sharing any transcript with an AI provider. Ganim's assessment guide expands on his discovery and report format.

Original assessment prompt

Draft a one-workflow assessment

Use approved notes and leave implementation for review.

Ready to copy

When the Assessment Leads to Recurring Work

Start with a referral, partner or business owner who opts into a short discovery call. Ganim mentions Treg, a pay-per-call catalog, as optional prospect research tooling. It isn't one of the three assessment repos and you don't need it to sell the work. Offer one useful observation on the call, then quote the full report separately. Agree on what a client will receive, what data they must supply and what is out of scope.

Four sales at $999 would be $3,996 in gross project revenue. Four separate $300/month monitoring agreements would add $1,200 per month in gross recurring revenue while all four remain active. Those are Ganim's scenarios, not sales results or profit estimates. Price in discovery, data cleanup, report review, follow-up, infrastructure and usage. A retainer should name what you actually do each month, such as rerunning a fixed prompt panel, verifying exceptions and reviewing one proposed change. A one-time fix can simply be a one-time project.

Video Chapters

TimeTopic
00:00Introduction: the $999 assessment offer
01:01ActivityWatch: find where work time goes
05:01Elmo: measure AI search visibility
07:43Browser Use: test the inquiry journey
11:41No-code AI tools assessment
12:37Free training workshop announcement

Sources and Limits

The video and supplied transcript are the source for Ganim's offers and workshop mention; see the original announcement for event details. Repository and vendor links above support the current feature and pricing claims. The pilot is an editorial adaptation, not a tested engagement. We did not verify sales claims, workshop availability or client outcomes. Evidence cutoff: 18 November 2026.

Common questions

What does the proposed $999 AI assessment include?
A scoped diagnostic with evidence, a recommended fix, and a review conversation. Implementation and ongoing monitoring are separate. The price is Corey Ganim's example, not a guaranteed sale.
Does ActivityWatch record video of the screen?
Its standard watchers record data such as active app, window title, browser tab metadata and away-from-keyboard status. This is not a screen video, but titles and URLs can still reveal sensitive information.
Which offer is easiest to start without technical setup?
The interview-based assessment. Map one workflow with a business owner, compare a small number of fixes and propose a measurable test before installing anything.
Share
X LinkedIn Reddit
Build Yours

Want a system
like this one?

Book a free 30-minute call. We map your situation, identify the highest-impact automation, and figure out if we are a fit.

Book Free 30-min Call