HardTally
Start free discovery
Menu

The GenAI program platform for companies rolling out AI across several departments.

Turn AI experiments into measured results.

When leadership asks what the company is doing with GenAI and what it has changed, answer from one screen: every initiative, its stage, its measured result.

Map AI opportunities in one department, free

A free Department Discovery: the department's workflows, where AI fits, a shortlist. No card.

Already tracking AI work in a spreadsheet? Import it
  • Free discovery, no card
  • The trial starts only when you log your first initiative
  • One flat company price, no seats
HardTallyThis weekAcme
  • 2 advanced
  • 2 results recorded
  • 1 copied by another team

  • Support · Ticket triage

    Value verified

    Tom Reyes

    Handle time 45 15 min/ticket

    Verified · high evidence · improvement

  • HR · Candidate screening

    Experiment

    Lena Brandt

    Time to shortlist 12 14 days

    Verified · medium evidence · worsening

  • Finance · Invoice coding

    Experiment

    Sam Okafor

    Measurement plan locked

    No result yet

  • Marketing · Inbound request triage

    Idea

    Jonas Weber

    Copied from Support · Ticket triage

    Not yet measured
Sample data · the screen your Monday meeting runs from

The problem

AI use is everywhere. Measured results are not.

GenAI work is happening in every department. The record of it lives in people's heads, inboxes and a spreadsheet nobody fully trusts, and the results get rebuilt from memory the day leadership asks.

  • Value reconstructed later

    Nobody wrote down the “before,” so the result is rebuilt from memory when leadership asks. The spreadsheet holds ideas, not measurements.

  • Follow-through in one head

    Owners, blockers and next steps live in one person's memory and inbox, whoever that is. When they are out, the program stands still.

  • Activity reported as outcome

    A faster draft gets reported as lower cost. “Support rolled out AI this week” gets reported as progress. Neither is a measured change in a business number.

  • Work repeated, progress invisible

    Nobody can see what the company is already doing with GenAI, so teams start from a blank page and the program looks stalled.

This is how AI programs lose their mandate.

Not because the experiments failed. Because when leadership asked, nobody could show what they did. The budget line moves, the weekly forum stops, and the two things that actually worked are never copied by the teams that needed them.

The usual fixes don't fix it.

  • The spreadsheet

    Holds ideas, not measurements. Updated for a month, then not.

  • The project portfolio tool

    Built for projects with budgets and milestones. Nobody logs a two-week prompt experiment in it, so it stays empty.

  • The vendor dashboard

    Counts prompts, seats and logins and calls it value. The vendor vouches for its own number.

How it works

One record, from first idea to measured result.

Three questions — where could AI help, what's already running, what changed — answered in one place, in this order.

  1. Capture

    Save an idea with a title and a problem. Rough estimates are fine.

  2. Put it in the company portfolio

    Every team sees it; any team can copy it, source credited.

  3. Agree the measure before work starts

    One metric, one baseline, agreed when the experiment begins. An estimate is enough to start.

  4. Record before and after

    Better, worse or neutral. Every result counts.

  5. The owner confirms it

    The person accountable for the metric stands behind the result.

Missing information blocks a claim of verification. It never blocks capturing an idea.

Already have candidates? Import your trackerImported work starts as new initiatives, never as verified history, and stays provisional until a measure is agreed.

Where could AI help?

Find where AI fits the work a department already does.

A Department Discovery maps one department's real workflows — described by the people who do them, or confirmed from the process documents and notes they already have — and suggests where AI could help, with the source for every suggestion. One goal — cost, cycle time, quality, throughput, revenue or customer experience — and one workflow to start. No workshop.

  • Every suggestion cites its local source. Where there isn't one, it's labeled “AI inference, no direct source.”
  • Potential and readiness are scored separately. An unanswered question stays unknown, never zero.
  • Nothing becomes an initiative until a person promotes it.
HardTallyDepartment Discovery · SupportAcme

Confirmed workflow

Investigating customer tickets

We review the ticket history, find the relevant policy, and gather context before drafting a response.

Local source: team workflow description

Suggested opportunity

Prepare the context.Keep the human judgment.

Assist the team with ticket summaries and retrieval of relevant policy passages before review.

Potential

83/100

Readiness

Unknown

4 of 7 answers

Why this fits

The confirmed workflow includes repeated information gathering. A summary may reduce preparation effort; its accuracy still needs testing.

Local source

The team's own description above. The suggestion is not based on a department name alone.

Prerequisites and risks

Permission to access ticket records, retrievable policies, a named reviewer, and a test for missed or incorrect context. Customer data requires appropriate handling.

Hypothesis only · No realized value claimed

  1. Describe
  2. Confirm
  3. Review
  4. Shortlist
Sample data

What's already running?

See what every team is doing. Copy what already works.

Program Leads and Department Heads see every department's initiatives, owners, stages and published results. Sensitive discovery files and raw survey answers stay with the department that owns them.

HardTallyCompany portfolioAcme

  • Support

    Ticket triage

    Tom Reyes

    Value verified

    Handle time 45 15 min/ticket

    Verified · high evidence · improvement

    Copy into my department

  • HR

    Candidate screening

    Lena Brandt

    Experiment

    Time to shortlist 12 14 days

    Verified · medium evidence · worsening

  • Finance

    Invoice coding

    Sam Okafor

    Experiment

    Measurement plan locked

    No result yet

  • Marketing

    Inbound request triage

    Jonas Weber

    Idea

    Copied from Support · Ticket triage

    Not yet measured

+4 more initiatives

Copies start as a new idea, source credited. Results are not carried over as verified.

This month

Ahead
Support 5
Keeping pace
Finance 1, HR 1
Quieter than peers
IT, Marketing, Operations

Counts of movement and results. No points, badges, or rankings.

Sample data

What did it change?

See the value AI actually created, and what it's based on.

Every initiative that reaches measurement ends in a result: one metric, measured before and after, with its source and its owner attached. Favorable, neutral or adverse, so the value you report is the value that happened.

HardTallyValue ledger · SupportAcme

Support · Ticket triagev1

Workflow metric · Handle time

Baseline

45 min/ticket

Result

15 min/ticket

Measured change−30 min/ticket · −66.7%improvement

Verification
Verified
Evidence confidence
High
Confirmed by
Priya Nair
Result period
Sep 14–20, 2026
Inspect the evidence
Baseline source
Helpdesk report, May 2026, all tickets · measured · from a system record
Result source
Helpdesk report, Sep 14–20, 2026, all tickets · measured · from a system record
Plan locked
Jun 1, 2026
Confirmed
Priya Nair · Sep 23, 2026
Version
v1

Measurement level

  • Workflow metric · Handle time

    45 15 min/ticket

    −30 min/ticket

    Verified · high evidence

  • Business outcome · Cost to serve

    EUR 4.10 EUR 3.60 per ticket

    EUR −0.50 per ticket

    Awaiting confirmation · medium evidence

The time measurement does not generate the financial result. Each has its own source and its own owner.

Sample data

What “verified” means here

  • One before-and-after, each figure with its own inspectable source.
  • Every result has an owner. The person accountable for the metric — Finance for cost, the department head for a workflow metric — confirms it.
  • Money is measured from a financial record, never derived from hours saved.
  • Adverse results are verified the same way as favorable ones.

Evidence confidence

Low
Either figure is estimated or self-reported, the metric was chosen after the fact, required evidence is missing, or the before and after can't yet be compared fairly.
Medium
Any other complete measured comparison, such as a documented human measurement, or a record-backed result still awaiting confirmation.
High
Both figures are measured and come from inspectable system, financial, or operational records, the owner has confirmed them, and no other known claim to the same change is still open.

The owner marks the result Looks right or Doesn't look right. No legal sign-off needed.

How verification works

An initiative moves through these stages:

  1. Idea
  2. Qualified
  3. Experiment
  4. Validated
  5. Implementing
  6. Adopted
  7. Value verified
  8. Scaled

Gate 1: starting an experiment requires a measurement plan. An explicit estimate is enough to start.

Gate 2: Value verified requires a before-and-after comparison with its evidence, confirmed by its owner.

Verified means currently confirmed: if the evidence is corrected or lost, the badge comes off, and a verified result can be favorable, neutral, or adverse.

What needs me this week?

Every Monday: what moved, what needs a decision, and the same week for leadership by email.See the weekly rhythm

The difference

The spreadsheet lists ideas. Telemetry counts activity. The ledger records results.

HardTally works beside your AI tools and doesn't monitor them.

Compared onSpreadsheet or NotionProject portfolio toolVendor AI telemetryHardTally
Unit of recordIdeas and statusProjects, budgets, milestonesPrompts, seats, loginsOne result for one metric over one period
What it showsA listPlans and datesUsageBefore and after, against a baseline agreed when work started
Who vouchesNobodyThe project managerThe vendorThe person accountable for the metric
Bad resultsDeleted or forgottenRarely enteredNot a conceptRecorded, verified and reported under their own heading
Reuse across departmentsTeams can't see each otherSeparate projectsOne tool at a timeOne portfolio; copy any initiative, source credited
Keeping it currentDies in month twoProject administrationAutomatic, and meaninglessAn idea needs a title and a problem

Why now

Why this week, not next quarter.

A baseline can only be recorded before the change. Every pilot that starts without one is a result you'll have to reconstruct later, and everyone will know it was reconstructed. The next experiment your teams start is the cheapest one to get right.

Pricing

One company price. No seats, ever.

Leads, Department Heads, contributors, survey respondents and the people who confirm results are never billed. Add every department; the price does not move. All prices exclude VAT.

  1. Free discovery

    Free

    No card, no clock

    • Unlimited saved Department Discoveries
    • Work map, suggestions and shortlist
    • Documents and a short fixed survey
    • Prepare direct-start drafts
    Map AI opportunities in one department, free

  2. 28-day trial

    28 days

    Once per company

    • Starts only when you log your first initiative
    • Full ledger, cockpit, verification and digest
    Start with your own initiatives

  3. Company plan

    EUR 490

    per month

    or EUR 4,900 per year

    One flat company price

    • Everything in the trial, continued
    • Card checkout or invoice
    • Upgrade inside HardTally, during the trial (unused days kept) or after it

If you stop, discovery stays usable, records stay readable and exportable, and nothing is deleted because a trial ends.

Guided Discovery Session — EUR 490 one-off, excl. VAT

Same product, same 28-day trial; the session is a service, not software access. Credited against the annual plan.

  1. Short prep
  2. 90-minute virtual session with you and one Department Head
  3. Written summary in your account
  4. 30-minute results check before trial day 28

The session ends with a shortlist of 2–3 opportunities, named owners and a first measurement plan. Start the 28-day trial at the end of the session, or later.

Book a Guided Discovery Session

What the first four weeks can look like

A simple way to use the trial. Your teams set the pace; the platform adds no waiting time.

  1. Week 1 — Discovery

    One department, one goal, one workflow. Confirm the work map, review the suggestions, shortlist two or three.

  2. Week 2 — First initiatives

    Promote the shortlist, name owners, agree one metric and one baseline each.

  3. Week 3 — Running

    Record what's happening as it moves; other departments see the portfolio and copy what fits.

  4. Week 4 — First results

    Record before and after for anything that has finished its measurement period and have the owner confirm it. Send the first weekly digest to leadership; a week with no verified result still shows progress.

Method Library

Field guides, templates and teardowns.

Read the 1 article

Get started

Start with one department.

One goal and one workflow are enough to begin. Lock the “before” for the next experiment your teams start.

Map AI opportunities in one department, freeImport your AI tracker

No card for discovery. Nothing is deleted when a trial ends.