Stop paying for hypotheses. Ship against evidence.
Jottings installs on consenting employees' Macs and PCs for two to four weeks, quietly measures where work actually goes, and hands you a ranked report of automation opportunities - with dollar impact, complexity, and the tool stack to build them in.
Ten AI automation opportunities, ranked by dollar impact.
Everyone is guessing. Expensively.
Consultants show up with hypotheses.
But you are paying for evidence. A PDF produced from six Zoom calls is not a diagnosis - it is a vibe, priced like a surgery.
Agents need context and evals.
Every serious AI job post now requires both. Nobody has the input data to produce either. That is the gap between a demo and a deployment.
Your team is staring at 400 workflows.
Which three will move the CFO's number this quarter? You can workshop it for six months, or you can measure it in three weeks.
Jot·tings
/ˈdʒɒtɪŋz/Short, honest notes made in the moment about what is actually happening - not what a manager says is happening, and not what a deck claims is happening.
The product is named after the unit. Thousands of small, redacted observations about real work, compiled into the one document nobody in enterprise AI currently has: proof of what to automate, in what order, for how much money, using which tool.
Jottings.AI is not a workshop, not a deck, not a vibes-based audit. It is the observation layer underneath every serious AI rollout - the place where context and evals actually come from.
Install. Observe. Synthesize. Report. Purge.
A bounded, consented engagement with a beginning and a hard end date. You get a document. The agent removes itself.
Install
Consultant or internal lead pushes the Jottings desktop app to consenting employees. Code-signed for Mac and Windows. A small menu bar icon lives in plain view. Pause or opt out any time, no consequence.
Observe
For two to four weeks the app passively notes which apps are open, for how long, and which sequences repeat. All PII is redacted on the device - regex, on-device NER, OCR redaction - before anything leaves the machine. Raw pixels never upload.
Synthesize
Thousands of redacted observations get fingerprinted, clustered, and reconstructed into a work graph. The system matches them against a growing library of Skill Cards - vetted automation patterns with tool stack, complexity, and expected ROI.
Report
A branded, editable PDF: top ten opportunities ranked by dollar impact, evidence links for each, complexity score, recommended tools, AI-readiness gap analysis, and a roadmap. Present it Monday.
Purge
The agent auto-uninstalls on the engagement end date. Raw engagement data is purged within 90 days. You keep the report, the customer keeps their privacy.
The one document nobody else will hand your CFO.
Every scan produces the same format. The cover, logo, and color palette swap for each consultant or team. Under ten minutes of editing.
- $ impact at a glance
- Ranked by quarter-one payback
- Signed by the attesting consultant
- Every recurring process, typed
- Department heat map
- Time-in-app by role, redacted
- Dollar impact + complexity score
- Recommended tool stack per card
- Evidence link to every claim
- Context packs required
- Eval harness template
- Change-management risk notes
Ten AI automation opportunities, ranked by dollar impact.
Every scan produces this. In under 10 minutes of editing.
Every scan makes every future scan smarter.
Redacted findings are k-anonymized across five or more organizations, then fed into the Skill Dojo - the growing public-but-curated library of Skill Cards. Each card is a vetted automation pattern with the tool stack, prompts, evals, and measured ROI ranges. Tools referenced in your report can carry your affiliate links.
What Jottings is not.
We do not replace n8n, Make, Lindy, or Relevance. We tell you what to build inside them.
HubSpot exists. Salesforce exists. We are not trying to be them.
Clay and Apollo exist. We enrich them with real signal from the work graph. We do not replace them.
Engagements are two to four weeks, tri-party attested, and auto-uninstall on the end date.
Not RescueTime. Not Hubstaff. We do not measure individuals - we measure workflows.
One product. Two people who need it yesterday.
Walk into any 50–500 person prospect, run a real 2–4 week observation, walk out with a closeable deal. No more 'where did these numbers come from.'
Branded reports. Your logo, your palette, your methodology. Tools referenced in the report carry your affiliate links - a new revenue line per engagement.
Growth add-on exports anonymized case studies and a lookalike prospect list directly into Clay, Apollo, or HubSpot.
“We stopped competing on slide design the day our report showed the CFO the actual dollar number.”
- Tri-party DPA
- State-aware monitoring notices (NY, CT, DE, and more)
- Local-first PII redaction
- Engagement-scoped encryption keys
- 90-day engagement data purge
- Employee opt-out without consequence
- Cyber and E&O insured
- SOC 2 Type I in progress
Pay for the engagement, not the seat.
Per-employee pricing punishes the thing you want more of. We price the scan.
- 1 active engagement
- Standard report branding
- All Skill Cards
- Email support
- 3 active engagements
- Full white-label branding
- Affiliate link injection
- Priority email support
- 10 active engagements
- Team seats and role permissions
- Priority report tuning
- Slack channel with our eng team
- Unlimited internal engagements
- Outcome evals dashboard
- Context pack exports
- Named customer engineer
Lookalike Intelligence, Case Study Generator, push to Clay / Apollo / HubSpot.
For overflow engagements beyond your plan's active count.
Bring Jottings inside. We do a short qualification call, then onboard your first business unit within two weeks.
Company tier is $15–50k/year, includes unlimited internal engagements, the outcome-evals dashboard, context pack exports, and a named customer engineer. We are taking a handful of design-partner teams this quarter.
Use a work email. We read every submission.
Your first scan tells you more than your last six months of workshops.
Install on one business unit. Two to four weeks later, present the report.
Code-signed · auto-uninstalls on end date · opt-out any time.