← Factory home Catalog Playbooks Honest

Playbook · Catalog editorial

Five patterns we keep seeing across 772 AI ideas

After scoring, ranking, and skeptic-memo'ing two-hundred-plus AI products, the same five shapes keep showing up. The patterns that actually work, the ones that fool you, and how to tell the difference.

When a catalog hits 772 products, you stop seeing ideas and start seeing shapes. The Factory has now scored, evaluated, and skeptic-memo'd enough AI product proposals to notice the underlying patterns. There are roughly five.

This essay is the editorial read on those five patterns. Each one names the shape, explains why it works in 2026, shows three to four products in the catalog that fit, and calls out the failure mode that fools founders into thinking they have the pattern when they don't.

If you're trying to decide which idea to unlock, this is the meta-layer above the catalog. Find the pattern that fits your situation, then pick a product within it.

The categories matter less than the shapes. A "legal" idea and a "trades" idea can be the same pattern (after-hours catcher). A "sales" idea and a "compliance" idea can be the same pattern (workflow compression). Read by shape, not by vertical.
01

Workflow compression: 90 minutes becomes 5 minutes

Shape: take a specific task that takes a senior 60-120 minutes today. Use AI to do 80% of it in 5 minutes. Senior reviews and edits the last 20%.

This is the dominant pattern in the catalog. About 35-40% of products fit it. The shape works because a senior person's time is expensive (their billable rate is often $150-300/hour) and the task is repetitive enough that pattern-matching by a model works.

The reason it actually pays in 2026: inference costs are low enough that the AI portion of the workflow costs less than $1 per execution, while the senior's hour costs $150+. The compression math is real.

Examples in the catalog: Bookkeeper AI (monthly close from hours to minutes), Email Marketing AI (writing emails), Architect AI (system design docs), RFP/Tender Writer (proposal drafting).
The failure mode: founders pick a workflow that's only done a few times per year. The math only works at high frequency (weekly or daily) because the cost of building the workflow is fixed. If your buyer does the task 3 times a year, they'll pay $300 once, not $200/month forever.
02

After-hours catcher: catching what humans miss when offline

Shape: a workflow that fires when no human is on duty. Phone calls, intake forms, support tickets, anomaly alerts. The AI handles first response; humans pick up next business day.

This is the highest-conversion-rate pattern in the catalog. About 12-15% of products fit it. The reason it converts so well: the buyer can mentally calculate the lost revenue from missed after-hours moments and the AI math is a no-brainer against that loss.

The Counsel AI graduation (now intakecounsel.com) is in this pattern. Solo lawyers lose a $5,000-25,000 case every time a prospect calls at 7pm and gets voicemail. The product math works at $300/month because one captured case pays for the year.

Examples in the catalog: AfterHours (night desk for service businesses), RecoverCall (missed call recovery), Intake AI (after-hours intake), Counsel (lawfirm-ai).
The failure mode: picking an "after-hours" workflow where the lost-opportunity cost is unclear. If a missed event costs the buyer $20, the math doesn't work. The pattern only fires when the cost of each miss is $500+.
03

Second-opinion play: review and explain existing work

Shape: take something humans already produce (a contract, a sales call, a marketing campaign, a closed deal). Use AI to review it, score it, identify issues, suggest improvements.

About 15-20% of products fit this. The reason it works in 2026: AI has gotten good enough to surface real issues without false-positive spam. Buyers used to ignore AI quality checks because the signal-to-noise was awful. The 2026 versions actually flag the right things.

The pattern's economic moat is interesting. A buyer paying $400/month for a "review" product doesn't switch easily because the AI has learned their context. Switching costs compound.

Examples in the catalog: Win-Loss Analysis AI (post-deal review), Pipeline Report (deal-by-deal review), Discovery Call AI (post-call analysis), Code Review AI.
The failure mode: shipping a product that grades but doesn't tell you what to do next. The buyer cares about the action, not the grade. If your output is "your sales call was 7/10," the buyer churns. If your output is "your sales call missed the budget question; here's the email to send to follow up," the buyer keeps paying.
04

Pre-meeting prep: brief the human before the moment

Shape: a human is about to enter a high-stakes interaction (sales call, customer success meeting, board prep, interview). AI assembles the relevant context in a 1-page brief delivered 5 minutes before.

About 10-12% of products fit this. The reason it works: senior humans hate prep work but love being prepared. The AI does the work, the human gets the benefit, and the alternative (not being prepared) is professionally embarrassing in a way the buyer feels.

Pricing is unusual for this pattern. Buyers will pay $50-150 per high-stakes meeting (one-off) or $200-500/month for unlimited prep. The math is paid against the value of the meeting itself, not against time saved.

Examples in the catalog: Lead Intel AI (pre-call intel on the prospect), Prospect Enrichment AI (name + domain to full profile), Seller AI (closing co-pilot), Account-Based Sales AI.
The failure mode: generating prep that's too long. A senior buyer doesn't read a 5-page brief 5 minutes before a meeting. The discipline is brutal one-pagers. Most builders ship 5-page briefs and wonder why retention is bad.
05

Paperwork compression: forms, compliance, reports

Shape: a regulated or process-heavy workflow requires producing structured documents (RFP responses, compliance reports, immigration paperwork, audit prep). AI fills 80% of the document from inputs and conversation, human reviews and submits.

About 15-18% of products fit this. The reason it works: paperwork-heavy workflows have the highest hourly burn rate (lawyers, auditors, compliance officers) doing tasks that are 70-80% pattern-matching. Compression here is more valuable than in any other pattern because the input hours are so expensive.

The catch: regulated workflows have compliance considerations. A product that generates a 90% accurate compliance report is worse than one that generates 60% accuracy because the buyer trusts the 90% and ships errors. The pattern only works when the AI is calibrated to flag uncertainty explicitly.

Examples in the catalog: Immigration Law AI (paperwork-heavy), Audit AI (audit prep), Compliance Radar AI, RFP/Tender Writer, Roofing AI (estimates and reports).
The failure mode: shipping a paperwork product that hides its uncertainty. A buyer who gets burned by an "I think this is right" output once will never trust the product again. The discipline is making the AI's confidence visible to the user, even when it's low.

How to use this lens when reading the catalog

The catalog filters and audience pages help you find products by buyer type. The patterns above help you find products by product shape. Those are different cuts.

Three suggested moves:

  1. If you have an unfair advantage in a specific workflow (you're a senior bookkeeper, you've been in legal intake, you've run compliance at a bank): browse by workflow pattern in your area. The vertical category in the catalog will help. So will the dossier's "named buyer" section.
  2. If you have channel access but no domain (you have a vertical newsletter, agency, or community): browse by pattern, not by product. After-hours catcher and pre-meeting prep travel well across verticals because the underlying buyer pain is universal.
  3. If you're not sure where to start: read three dossiers in three different patterns. The one where you immediately think "oh, I know exactly who would buy this" is your answer.

The patterns that did NOT make this list

Three shapes show up in the catalog repeatedly but did not make the cut as patterns we'd recommend pursuing:

These appear in the catalog because they get proposed often. They are not in the curated audience pages or the playbooks because the operator math does not work for self-funded plays.

The next step

Pick a pattern that matches your situation. Browse the catalog within that pattern. Read 2-3 dossiers in your top picks. Unlock one for $5.

If you'd rather have us pattern-match the right product to your situation, the operator partnership Phase A conversation ($2,500) is structured around exactly that. We read the dossiers with you and pick the one that fits your unfair advantage.

Read next.

Market thesis

Why vertical AI eats horizontal AI in 2026

The market thesis behind the catalog's category choices.

Scoring rubric

How to read an Adoptability score honestly

The 10-axis score on every product page, in operator-honest terms.

All 11 essays at /factory/playbooks/.