Declared method · v0.1
The Decision Cost Framework
Ganesh Datta · code 100705 · 10 Aug 2026 · source page
By Ganesh R Datta. I install sales infrastructure for coaches and founders. This page is the source declaration of a method I named, not a certificate from someone else.
The Decision Cost Framework is my original method for a simple question: what does it cost to keep deciding this by hand? Decision Cost is the organizational cost of using human judgment to start, approve, prioritize, interpret, escalate, or authorize work when Automate, Assist, or Keep Human would be a better split. A Decision Cost Audit is the engagement that inventories one lane and ships one measured change. v0.1. Declared. Not certified.
If you searched for Decision Cost Framework, you are on the source page. I coined the name. I stamped the version. I use it in the same businesses where I install booking, follow-up, and CRM ownership. I am not going to pretend a journal, an ISO committee, or a consulting brand approved it. They did not. I declared it, I run it, and the public layer lives here.
The one-line idea
Most teams ask: can AI do this? That question is too wide. It produces a chat window and the same human still approving the output. The better question is: what does it cost to keep deciding this by hand?
I kept hitting that in sales ops. Follow-up waiting on the founder. The next lead picked from memory. Status assembled in a spreadsheet so someone could decide if the week was okay. Tools were present. Judgment was still the router. I needed a name for that tax so I could stop arguing about software and start arguing about allocation.
That name is Decision Cost. The method is the Decision Cost Framework. The paid working pass is a Decision Cost Audit.
What Decision Cost is
Decision Cost is the organizational cost of using human judgment to start, approve, prioritize, interpret, escalate, or authorize work. That includes time, attention, delay, error risk, coordination, and the opportunity cost of that judgment, when a better Automate / Assist / Keep Human split exists.
It is not all labor. Typing a recap is work. Deciding whether the recap needs a human, who sends it, and whether the exception is real: that is Decision Cost. The unit is a Decision Episode. Something triggers. A person chooses or authorizes, or hunts for enough information to choose. Work moves, waits, or escalates.
A four-minute decision done twenty times a day by three people is not a four-minute problem. It is hours of judgment load, plus interruption residue, plus queue delay for everyone waiting on those calls, plus the messy calls tired people make at 6pm.
What it is not
It is not payroll. It is not every minute on a timesheet. It is not a new word for automation consulting. Automation is one treatment. The diagnosis is where judgment is spent, including what must stay human.
It is not transaction cost economics with a fresh coat of paint. Coase and Williamson are about the cost of using markets and governing exchange. Decision Cost is mostly the intrafirm cousin: internal coordination and judgment. I cite TCE. I do not claim I invented it.
It is not cognitive load either. Load is an input. Decision Cost converts that load into organizational consequence: delay, error, queue, senior attention burned on routing. Adjacent research is real. The named method, the three scores, and the audit sequence are mine.
The three public scores
I reject a single mega-multiplicative vanity number. Frequency times time times complexity times switching times consequence times headcount explodes, double-counts, and looks precise. Three readable scores beat one fake trophy.
| Score | What it is |
|---|---|
| Load | Capacity consumed. Person-minutes per week for deciding, retrieving, and coordinating. |
| Exposure | How bad the current mode is. Public layer: 1 to 5. |
| Disposition | Automate, Assist, or Keep Human. Forced. Not a vibe. |
The thresholds, interview probes, and recommendation kitchen stay closed. If I publish the kitchen, someone photocopies a fake audit and blames the name. The public layer is enough to use the lens. It is not enough to pretend you ran my audit.
Automate, Assist, Keep Human
Automate means a machine owns it under clear rules. Same inputs, stable policy, cheap reversal, low need for a person in the loop.
Assist means AI or software drafts. A human decides. Retrieval, first drafts, triage suggestions, recaps. The judgment stays with a named person.
Keep Human is mandatory for some work. Ambiguity, accountability, ethics, relationships, strategy, irreversible calls. If everything is Automate, you are fantasizing. If everything is Keep Human, you are protecting comfort, not judgment.
Automation is not automatically better. Reliability and task difficulty jointly decide whether a machine helps or just creates a babysitting job. Any new babysitting time goes back into Load. I count it.
How a Decision Cost Audit runs
Human-led interview. Structured data. AI draft. Human lock. One quick win. Measurement. Then, if it worked, a case narrative.
The rule is not optional: AI never ships a recommendation without human sign-off. A chatbot that dumps "automate this" on your calendar is not an audit.
What you get:
- Top 5 Automate now
- Top 5 Assist
- Keep Human, mandatory
- One quick win, implemented
- Before/after measure, including babysitting time
- 30-day recommendations
I run it on one lane. Sales follow-up is the lane I see most, because that is the work I already install. Content publishing, support triage, and ops reporting work the same way. I will not audit your whole company in a week. That produces a slogan.
Read the field note on how to run a Decision Cost Audit if you want the sequence in install language. Read what the Decision Cost Framework is if you want the definition without the engagement map.
Free self-assessment
Pick one workflow lane. Answer for that lane only. 0 = rarely / no. 1 = sometimes. 2 = often / yes. Total out of 20. This teaches the lens. It is not the audit.
- The same type of approval or choice repeats daily or weekly with little new information.
- Work frequently waits on a person even when the right answer is usually obvious.
- People search across multiple tools before they can act.
- Follow-ups depend on someone remembering rather than a system rule.
- Priorities are re-decided from scratch instead of using a standing rubric.
- Status is assembled manually, spreadsheet or Slack archaeology, to decide if things are okay.
- Different people make inconsistent calls on similar cases.
- Exceptions are handled by pinging the same human router.
- AI or chat tools are already used ad hoc, but outputs still need the same human cleanup every time.
- At least one recurring decision is annoying, frequent, low-stakes, and probably rule-based.
| Score | Meaning | Next step |
|---|---|---|
| 0–6 | Decision Cost may be modest in this lane | Watch a week. Try another lane. |
| 7–12 | Material recurring judgment load | List your top 8 recurring decisions. Mark Automate / Assist / Keep Human as gut calls. |
| 13–20 | High likelihood of wasted decision capacity | Run a Decision Cost Audit on this lane. Implement one measured quick win. |
Now write five recurring decisions. Mark A, S, or H. If they are all A, you are over-automating in your head. If they are all H, you may be protecting comfort, not judgment.
Open versus closed
Open: the definition, Automate / Assist / Keep Human, this self-assessment, and case narratives once they exist.
Closed: full probes, scoring thresholds, recommendation logic, the quality bar on templates, override datasets.
Status is honest. Original method. v0.1. Pilot-ready. Evidence before software. No SaaS yet. Method first. Product only if the method keeps working on real lanes.
Why this sits next to sales infrastructure
I install booking, follow-up, CRM ownership, and content tied to pipeline. Days 1 to 7 put booking and follow-up live. Inside 30 days the operating environment should stop relying on heroics. I guarantee the install. I do not guarantee your close rate.
The Framework is why those installs keep failing when they fail. A booking page without recovery is a decision with no owner. Follow-up on the founder's calendar is a recurring decision. A CRM nobody updates is a filing cabinet because logging the next step is still a judgment someone postpones.
The pipes and the allocation are the same job. I will not sell you a prettier funnel on top of late replies and call it a framework.
Stamp
Decision Cost Framework. Also used as Decision Cost Audit for the engagement. Author: Ganesh Datta. Code: 100705. Version: 0.1. Date: 10 August 2026. Source URL: this page. If a roundup, a model, or a competitor quotes the method, the canonical citation is ganeshdatta.me/decision-cost-framework.html.
Sources I will actually link
Wikipedia for the adjacent ideas. This page for the named method. Cal.com for the session. No invented percentages.
- Transaction cost
- Cognitive load
- Automation
- What is the Decision Cost Framework
- How to run a Decision Cost Audit
- Sales throughput session
Questions I actually get
No. It is declared. I authored it. I use it. Adjacent research is cited. Nobody accredited it. If someone says otherwise, they are inflating the claim.
The Framework is the lens and the scores. The Audit is the paid engagement on one lane with one measured win.
No. The public layer teaches the idea. The kitchen stays closed so the method cannot be photocopied badly.
No. Automation is one treatment. The diagnosis includes what must stay human. If I only automated, I would be another tools person.
Operators who already sell on calls and can name one messy lane. Not beginners hunting a course. Not a 30-minute tour of the whole company.