Wrong project
Too broad, no data, or the fix is already known.
Taktstone: a 12-question GE-style screen recommends DMAIC, Kaizen, Just-do-it, redesign or No-go — before anyone is assigned.
AI agents for Lean Six Sigma
Finish every improvement project.
Your MBB coach is built in.
Taktstone picks the right project, coaches your team through every DMAIC step, explains the statistics in plain language and challenges you at each tollgate — the way an experienced MBB would. The project closes only when the result is proven with data and signed by its owners.
No sign-up for either. Three finished projects — manufacturing, IT and AI — open in your browser.
- Every number computed by a tested statistics engine — never typed by AI
- Coaching for manufacturing, IT digitalisation, AI/ML, healthcare and services
- Your data stays yours: no trackers, team workbench on EU-adequate hosting
12screening questions before you commit a team
5tollgates with role sign-offs and an MBB-style audit
14analyses explained in plain language
8specialist AI agents
0numbers made up by AI
Why improvement projects stall
The method is proven. The missing piece is a coach.
Lean Six Sigma has decades of tools behind it. Projects still fail for the same five reasons — and each one is something an experienced Master Black Belt would catch.
Analysis paralysis
The team doesn't know which tool comes next.
Taktstone: the Coach shows the next step for your phase, why it matters and opens the right tool.
Misread statistics
A p-value on a slide, nobody sure what it means.
Taktstone: every test, control chart and Gage R&R comes with a plain-language reading and "what to do".
Weak tollgates
Gates approved on opinions; solutions before root causes.
Taktstone: an automatic audit flags small samples, unchecked gauges, unverified causes and jumping to solutions.
Gains that fade
Six months later the process has drifted back.
Taktstone: closure requires ≥ 20 stable points after the change, a control plan with owners, and validated savings.
See it finished
Three complete DMAIC projects you can open right now
Worked demo projects with realistic raw data. Every before/after number below is computed by the Taktstone engine from that data. Companies and people are fictional.
Manufacturing
Injection-moulding scrap and lead time
- 6.98% → 2.21% scrap
- 0.73 → 1.46 Cpk
- 14.3 → 4.3 days order lead time
- $184.7k/yr validated savings
Gage R&R first, root causes proven with t-test, regression and χ², SMED and a dryer interlock, 30 stable points after the change.
Open this projectIT digitalisation
Accounts-payable invoice processing
- 9.07 → 1.88 days invoice turnaround
- 74% → 93% first-pass match
- $12.67 → $6.84 cost per invoice
- $874k/yr validated savings
Process mining of an ERP event log, exception Pareto, ANOVA by channel, auto-match pilot, Kanban roll-out.
Open this projectAI / ML delivery
Model-change lead time and failed releases
- 40.4 → 9.2 days lead time
- 30% → 6.7% failed releases
- 31% → 66% GPU utilisation
- $1.46M/yr validated savings
Value stream from Jira, ServiceNow, MLflow and ArgoCD events, a rejected-then-approved Measure gate, golden-path releases.
Open this projectWhat you get
A coach, a statistics engine and every Lean Six Sigma tool — connected
Methods pioneered in the Toyota Production System and in Motorola and GE quality programmes, rebuilt for factories, IT departments and AI teams.
Coach: what to do next
A phase roadmap that ticks itself off from your project data, a tollgate audit of the mistakes an MBB catches, and every analysis explained in plain language.
Follow-up that sticks
Tasks with owners and due dates, shared project notes, private notes, and a lessons-learned library so the next project starts smarter.
Is it worth solving?
A 12-question GE-style screen scores strategy, customer pain, money, data, root-cause uncertainty, scope and ownership — and recommends DMAIC, Kaizen, Just-do-it, DMADV or No-go.
Measurable STAR goals
Situation, Task, Action, Result — with automatic SMART checks, so every charter states a baseline, a target, a deadline and the business result.
Agents for every phase
A Master Black Belt reviewer, process analyst, RCA investigator, statistician and finance validator — each on the right model for the job.
Real statistics, no spreadsheets
Gage R&R, I-MR, Xbar-R and p-charts with Nelson rules, capability, t-tests, ANOVA, non-parametric tests, χ², regression and normality — straight from CSV exports.
Diagrams you can work in
Value stream maps, swimlane process maps, fishbone, branching 5 Whys, FMEA, waste heatmaps and RCA traceability — editable, zoomable, exportable.
Kanban and flow
Boards with WIP limits plus cumulative flow, cycle-time percentiles, throughput and WIP aging computed from every card move.
Tollgates and sign-off
Business owner, process owner, project owner, lead, coach and finance approve each gate. Approval is blocked while deliverables are open.
Automated goal acceptance
CTQ targets met, improvement sustained on a control chart, root cause verified, actions closed, savings validated — checked from data before anyone signs.
Chart Studio
Paste from Excel or Sheets and build Pareto, run, histogram, box, scatter, control-chart or heatmap views inside the project.
Why not just…
The coaching layer the other tools leave out
| General AI chat | Spreadsheets & statistics packages | Enterprise process mining | Taktstone | |
|---|---|---|---|---|
| Tells you the next DMAIC step | If you ask well | — | — | ✓ Coach roadmap per phase |
| Statistics you can defend | Can make numbers up | ✓ but you must interpret | Flow metrics only | ✓ computed by code, explained in words |
| Tollgates, sign-offs, audit trail | — | — | — | ✓ role sign-offs, append-only |
| Proof the gain holds | — | Manual | Partial | ✓ automatic goal acceptance |
| Time to first value | Minutes | Weeks of training | Months of integration | Minutes — try the demo |
Built for
Whoever is leading the project
Green & Black Belts
Run your project with an MBB looking over your shoulder: next steps, plain-language statistics and a pre-check before every gate.
CI & quality leaders
A portfolio of screened projects, consistent tollgates, validated savings and lessons that carry across teams and sites.
Consultants
One workspace per client, switch in a click. White-label deliverables and multiply the projects you can coach at once.
Process owners new to Lean Six Sigma
Start from a worked example in your industry and follow the steps — no certification needed to begin.
Regulated industries
Audit trail, append-only sign-offs, self-hosted or on-prem builds and a validation pack.
Trainers & academies
Hands-on labs with finished case studies your students can open, inspect and copy.
How it works
From idea to accepted result
- 1
Screen and charter
Answer the screening questions, write the STAR goal with worked examples for your industry, name owners and approvers.
- 2
Measure with your data
Drop in ERP, MES, ServiceNow, Jira or MLflow exports. Analyzers build the value stream, baseline sigma, control charts and Pareto.
- 3
Verify root causes
Fishbone, 5 Whys, waste walks and hypothesis tests — every cause verified with evidence and traced to an action and a control.
- 4
Improve, control, accept
Pilot, prove the shift on a control chart, validate savings with finance, pass the audited tollgates, capture lessons, close when acceptance is green.
Start free — Community edition
Get early access
The free Community edition installs in Claude Code in one minute: core DMAIC skills, fishbone, 5 Whys, Pareto and calculators. We are onboarding early users personally — ask for access and we will send the install link and a short walkthrough.
Requires Claude Code with a Claude Pro, Max, Team or Enterprise plan (billed by Anthropic) and Node.js 18+.
Pricing
Priced for practitioners, ready for enterprises
USD, billed annually unless noted. Every paid plan includes all industry packs and updates.
Community
Free
Core DMAIC skills, fishbone, 5 Whys, Pareto, sigma and OEE calculators.
- Open source
- 1 industry pack
Most popular
Pro
$149/yr
The full workbench for one practitioner.
- SPC, capability, Gage R&R
- VSM, FMEA, Kanban, Chart Studio
- Tollgates & goal acceptance
- All industry packs
Consultant
$490/yr
For independent Black Belts and boutique firms.
- White-label deliverables
- Unlimited client workspaces
- Commercial-use licence
Team
$39/user/mo
Hosted team workbench for CI and quality departments. Min. 5 seats.
- Portfolio & prioritisation
- Role sign-offs & audit log
- Onboarding session
Enterprise
$15k+/yr
Regulated & air-gapped: med-device, pharma, aerospace, banks, hospitals.
- On-prem, signed builds, SBOM
- Validation pack (+$7.5k)
- SLA & roadmap input
Prices exclude your Claude plan. Academic: free for students, $1,500/yr department licence — ask us.
Work with us
More ways to grow with Taktstone
Training providers & academies
Bundle hands-on AI labs into your Yellow, Green and Black Belt courses. Co-brand or white-label, instructor dashboard, exam-aligned case studies.
From $2,500/yr + $15 per student
Become an academy partnerPaid DMAIC sprint
Our Black Belt runs one of your projects end to end with your team using Taktstone — from screening to a signed acceptance record.
Fixed fee $7,500–$25,000
Book a sprintRegulated industries
Local-first by design: no third-party APIs in the workbench, audit trail, append-only sign-offs, and a validation pack for computer software assurance.
Design-partner pilots available
Request the packPrivacy by design
Your process data stays yours
- The agents run in Claude Code on your own Claude subscription — we never proxy or resell AI usage.
- The workbench makes no third-party calls: no analytics, no CDN, no tracking. Every number comes from a tested, local engine.
- Host the team workbench on your own server. Sign-in uses one-time email codes; every save and sign-off is audit-logged.
Content you ask Claude to read is processed by Anthropic under your Claude plan’s terms.
FAQ
Questions
Do I need an AI API key?
No. The agents run inside Claude Code using your own Claude subscription. The workbench itself uses no external APIs.
Does it replace a Master Black Belt?
It gives every project the structure, checks and explanations an MBB would bring — next steps, tollgate audits, plain-language statistics and proof of results. For high-stakes decisions or advanced studies, keep an expert reviewer; Taktstone makes their time go much further.
Do I still need a separate statistics package?
For most DMAIC work, no: control charts with run rules, capability, Gage R&R, t-tests, ANOVA, non-parametric tests, χ² and regression are built in. Designed experiments (DOE) are on the roadmap.
Is the AI making up my numbers?
No. Every statistic is computed by a tested engine from your raw data; the AI agents and the coach only explain and question the results.
Is my data used to train AI models?
We do not train models on your data. When you use the agents, Claude processes content under your own Claude plan and Anthropic's terms.
How is “goal acceptance” decided?
From data: every CTQ meets its STAR target, the improvement holds on a control chart, a root cause is verified, actions are closed, the control plan is complete and finance has validated savings. Then the named owners sign.
Can we host it ourselves?
Yes. The team workbench runs on a single Linux server behind HTTPS with email one-time-code sign-in, roles and an audit log. Docker and systemd setups are included.
What does the Founding Member offer include?
Pro at $99/yr instead of $149, locked for as long as you stay subscribed, for the first 500 customers.
Your next project, coached from screening to signed result
Open a finished project, score your own idea in two minutes — or walk through it on your process in a 30-minute call.