- Brand
- Kaizumi
- Category
- Operations
- Primary Subcategory
- ERP & Operations Resource Tools
Integration details
Description
Kaizumi Lean Tools turns what you describe into the charts and sheets continuous improvement teams use every day. Tell ChatGPT about your process, paste your data, or upload a spreadsheet, and Kaizumi draws the tool right in the chat. Problem solving: A3 report, 5 Whys, fishbone (Ishikawa) diagram, Pareto chart. Flow: value stream map (VSM), takt time, kanban sizing, printable kanban cards, SIPOC, EPEI. Standard work: time observation sheet, yamazumi operator balance chart, standardized work combination table, process capacity sheet, work instruction (TWI job breakdown). Equipment: OEE, SMED changeover analysis. Six Sigma: control chart (SPC), process capability (Cp, Cpk), FMEA, Gage R&R, DPMO and sigma level. Every card can be edited in place. Each one opens in the matching free tool on kaizumi.com for printing or saving. No account or sign-in is needed.
- Integration type
- Plugin
- Verification status
- Not applicable
- Platform
- ChatGPT
- Primary Subcategory
- ERP & Operations Resource Tools
- Secondary Subcategories
- None listed
- Brand
- Kaizumi
- Access
- No account required
- First tracked
- 2026-10-09
- Tool count
- 23
- Geography
- US
The Primary Subcategory used for this profile’s headline score.
Other Subcategories where the Integration is listed.
Agent Discoverability
ChatGPT Plugin Discovery ScoreYour score is coming
ChatGPT now suggests Plugins on its own when they match a user's request.Your Plugin Discovery Score measures how often yours appears, and it will show here as soon as it’s ready.
What discovery looks like
Get alerts for Kaizumi Lean Tools
Get updates when Kaizumi Lean Tools’s Discoverability Score or category rank changes.
Competing in ChatGPT ERP & Operations Resource Tools
View Category23 tools agents can invoke
Use this when someone wants to build or update a 5 Whys causal chain from a problem and known explanations. Show a real visual chain and an evidence table. Accept zero, one, or several known answers; do not invent answers to reach five. Example: {"problem":"Wrong delivery","whys":[{"answer":"An old label was selected"},{"answer":"The old print job remained in the queue"}]}. Preserve the supplied order. Do not label the last answer a confirmed root cause unless the user explicitly confirms it. Do not invent countermeasures, blame, or evidence. After rendering, ask one question about the next missing link. On a follow-up, reuse the full previous input and append or correct only what the user supplied. Use this standalone tool unless the user asks for an A3. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
five_whys
Use this when someone wants to write up a problem as an A3 or a one-page problem-solving report, or when they talk through a problem loosely: what is going wrong, numbers, why, what they plan to change, who will do it. Build a partial report immediately with concise narrative AND appropriate populated visual blocks. Infer the blocks from the data: steps/handoffs → process, defect or loss breakdown → pareto, category measurements → bar, dated measurements → line, mixed observations → table, stated causes → fishbone or fiveWhys, agreed work → actions, measured before/after → comparison. Do not ask the user to choose a visual or enter data manually when their words already supply it. A single section may contain several complementary blocks. Respect requests for text only. Leave missing sections empty and ask one focused question about the next gap. Never invent measurements, causes, owners, deadlines, or completed results; label unverified causes as hypotheses. To continue, reuse reportId and pass the latest full content as base (prefer the widget draft context if edited), then send only changed text and new or replacement blocks. Reuse block IDs for replacements. Omitted fields are preserved; an empty string clears a text field. Returns an editable A3 and the next missing section. Only put user-supplied facts, hypotheses, targets, and agreed work in the report. Do not fill empty sections with your own hypotheses, proposed targets, or measurement plans. Offer suggestions in chat for the user to accept first. BLOCK FORMAT: each block must include section and type; other fields are flat, without a data wrapper. Exact examples: {"type":"pareto","section":"currentCondition","unit":"errors","items":[{"k":"Wrong item","v":48},{"k":"Missing item","v":27}]}; {"type":"line","section":"currentCondition","caption":"Late orders per day","series":[{"k":"Mon","v":12},{"k":"Tue","v":9}]}; {"type":"process","section":"currentCondition","steps":[{"k":"Intake"},{"k":"Packing","problem":"Queue"}]}; {"type":"table","section":"currentCondition","head":["Observation","Finding"],"rows":[["Packing","Queue"]]}; {"type":"actions","section":"plan","rows":[{"action":"Trial checklist","due":"Next week"}]}. Use k/v for chart points, not label/value. Use caption for a figure title. Do not copy example facts into the user report. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
a3_report
Use this when someone wants a control chart (SPC chart) to see whether variation over time is routine or a sign that something changed. Build an I-MR chart for one measurement at a time, an X-bar/R chart for subgroups of 2-10 measurements, or a p-chart for defectives counted out of inspected samples. Show the center line, control limits, and flagged rule violations. Example: {"chartType":"imr","title":"Fill weight","unit":"g","values":[502,498,501,499,503,500,497,502]}. Data must be in time order: I-MR needs at least 5 measurements, X-bar/R at least 4 subgroups of the same size, and a p-chart at least 5 samples. Do not invent, reorder, or drop data, and ask for missing values. Do not call the process capable, and do not call it in control beyond what the three rules show. Control limits are not specification limits. On corrections reuse the full previous input and change only the requested values. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
control_chart
Use this when someone wants a process sigma level, DPMO (defects per million opportunities), yield, or DPU (defects per unit) from units inspected, opportunities per unit, and defects found. Example: {"processName":"PCB Line 3 final inspection","units":1500,"opportunitiesPerUnit":7,"defects":63}. Ask for any missing count; do not invent units, opportunities, or defects. Count only the ways a unit can fail that a customer would call a defect. The sigma level adds the conventional 1.5σ shift, as on the site; say so when you report it. Defects cannot exceed units × opportunities per unit. Zero defects reports 6.0+σ. On corrections reuse the full previous input and change only the requested value. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
dpmo_sigma
Use this when someone wants to calculate EPEI (Every Part Every Interval), find how often a shared machine or cell can cycle through every part, or size lots per interval from part demand, cycle time, changeover time, and available minutes per day. Render the calculation, a parts table with run minutes and lot per EPEI, and a chart of how the day is used. Example: {"availableMinutes":420,"parts":[{"name":"A-104","dailyDemand":120,"cycleTimeSec":42,"setupMinutes":18},{"name":"A-108","dailyDemand":80,"cycleTimeSec":55,"setupMinutes":20}]}. Ask for missing available minutes, demand, cycle time, or setup time; never invent or default them, and never assume a shift length. Include every part that shares the process. When run time uses all available minutes, the tool reports over capacity with no EPEI; say so plainly rather than inventing a number. On corrections reuse the full previous input and change only the requested value. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
epei_calculator
Use this when someone wants an FMEA (failure mode and effects analysis): a list of ways a process step can fail, each with its effect, cause, current controls, and Severity, Occurrence, and Detection ratings from 1 to 10. The tool multiplies S × O × D into the RPN, sorts rows from highest RPN, and marks risk as high (severity 9–10 or RPN above 200), medium (RPN 101–200), or low. It shows the table, a Pareto of RPN by failure mode, and an optional after-action RPN. Example: {"process":"Bracket welding cell","rows":[{"step":"Weld cycle","failureMode":"Wire feed stalls mid-weld","effect":"Incomplete weld","cause":"Liner wear not tracked","controls":"None","severity":9,"occurrence":4,"detection":8,"action":"Replace liner on PM schedule","owner":"Maintenance"}]}. Do not invent failure modes, effects, causes, ratings, actions, or owners. Leave out any rating the user has not given; the row shows as not yet scored, and you ask for the missing rating. Detection is inverted: 1 means current controls will certainly catch the failure, 10 means they cannot. On corrections, reuse the full previous input, keep each row id, and change only the requested values. Omit id for a new row. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
fmea
Use this when someone wants to organize possible causes of a problem, draw a fishbone/Ishikawa diagram, or group causes by Method, Machine, Material, People, or other useful categories. Render the actual diagram using only causes supplied in the conversation. Unknown causes stay empty; do not fill a six-category template with guesses. Default every cause to hypothesis. Only mark supported when the user supplies evidence, and include that evidence. Example: {"problem":"Dough collapses","categories":[{"name":"Method","causes":[{"text":"Long fermentation"}]}]}. Ask one focused investigation question. For a correction or added cause, reuse the complete previous input and change only the requested part. Use this standalone tool unless an A3 is requested. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
fishbone_diagram
Use this when someone has run a crossed Gage R&R study (every operator measures every part more than once) and wants %GRR, repeatability (EV), reproducibility (AV), part-to-part variation, the number of distinct categories (ndc), or whether the measurement system is acceptable. The tool uses the AIAG average-and-range method, the same math as the Kaizumi Gage R&R calculator. Example: {"unit":"mm","tolerance":0.5,"parts":[{"operatorA":[20.02,20.04],"operatorB":[20.06,20.05]},{"operatorA":[20.11,20.13],"operatorB":[20.15,20.14]}]}. Each part holds each operator's trials in order: 2 to 10 parts, 2 or 3 operators (operatorA, operatorB, optional operatorC on every part), and 2 or 3 trials, the same count everywhere. Do not invent, average, round, or fill in readings; ask for any missing reading. Supply tolerance (upper minus lower spec limit) only when the user gives it; never guess it. %GRR under 10% is acceptable, 10% to 30% is conditionally acceptable, and over 30% is unacceptable; ndc must be 5 or more. On corrections reuse the full previous input and change only the requested values. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
gage_rr
Use this when someone asks what Kaizumi can do, which lean or Six Sigma tools are available, or wants to browse the tools. Shows a menu card grouped by purpose; each tile starts that tool in the chat. Do not use it when the person has already named a tool or supplied data for one; call that tool directly. The card already lists every tool, so do not repeat the list in your reply; add one short sentence inviting them to pick a tool or describe their problem.
kaizumi_tools
Use this when someone wants to size a replenishment loop, calculate kanban cards, or find loop inventory from demand, replenishment lead time, container quantity, and a chosen safety buffer. Show the calculation and a graphic separating demand coverage, safety buffer, and whole-container rounding. Example: {"dailyDemand":480,"leadTimeDays":0.5,"containerQuantity":40,"safetyFactorPct":15}. Ask for a missing safety factor; never silently assume a default, including zero. The same applies to missing demand, lead time, or container size. Demand and lead time must use the same working-day basis; clarify ambiguous hours. Do not claim the chosen buffer guarantees a service level. To print the cards for this loop, call kanban_cards with copies set to the card count. On corrections reuse the full previous input and change only the requested value. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
kanban_sizing
Use this when someone wants printable kanban cards: two-bin, production, withdrawal, supplier, or signal cards for one part or a whole list of parts, from a description, a pasted list, or an uploaded parts file. Draw every card right away from what the user gave, then ask one question about the most important missing detail. Example: {"environment":"manufacturing","cards":[{"type":"two-bin","name":"Hex head bolt, M8 x 25","code":"BLT-M8-025","quantity":100,"unit":"pieces","source":"Main stores, Rack A3","destination":"Assembly, Station 04","copies":2}]}. Use the item names, codes, quantities, and locations exactly as supplied. Never invent part numbers, quantities, or locations; leave them out and the card shows them as missing. Omit trigger and instructions to use the standard wording for the card type. Put a product or purchase link in qrContent so the QR code opens the reorder page; for supplier cards without one, ask for it. When the user sized the loop with kanban_sizing, set copies to that card count. Keep each text field short so it fits on the card. On corrections reuse the full previous input and change only the requested values. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
kanban_cards
Use this to calculate Overall Equipment Effectiveness from planned time, run time, ideal cycle time, and unit counts. Returns editable inputs and availability, performance, quality, and OEE percentages. Performance above 100% is flagged for input review, not silently capped. Edit inputs directly in the widget or through chat. For updates, reuse the latest complete input (including documentId), preferring the manually edited draft, and change only requested values. Start a new document without documentId. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
oee
Use this when someone supplies counts or loss amounts by category and wants to see the largest contributors, prioritize improvements, or create a Pareto chart without a full A3. Render the actual chart immediately from their data. Returns sorted bars, cumulative percentages, a data table, and the smallest group reaching 80%. Use items:[{"label":"Wrong items","value":48},{"label":"Missing items","value":27}] with a common unit. Ask for missing quantities rather than guessing. Repeated labels are combined. On a correction, reuse the previous complete input and change only the requested items; the tool is stateless. Do not call a3_report unless the user wants an A3. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
pareto_chart
Use this when someone wants a process capability study (Cp, Cpk) to see whether a process can meet its specification limits. Supply individual measurements in time order and at least one specification limit (lsl, usl, or both), plus an optional target. Returns Cp, Cpk, Cpu and Cpl, estimated PPM out of spec, predicted yield, sigma level (3 × Cpk), and a histogram with the fitted normal curve and the limits. Example: {"title":"Shaft diameter","unit":"mm","lsl":24.95,"usl":25.05,"measurements":[25.01,24.99,25.00,25.02,24.98,25.00,25.01,24.97]}. The tool needs at least 2 measurements; 30 or more is better. Do not invent, round, reorder, or drop measurements, and never invent specification limits or a target; ask for them. Sigma is the overall sample standard deviation, as on the site. Call the process capable only when Cpk is 1.33 or higher. Capability assumes a stable process: if stability is unknown, suggest control_chart on the same data first. On corrections reuse the full previous input and change only the requested values. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
process_capability
Use this when someone wants a process capacity sheet (Standardized Work Form 1), machine capacity per shift, or the bottleneck step from manual time, auto time, completion time, tool change interval and duration, and available time per shift. Show a capacity bar per step, the step table, and the bottleneck calculation. Example: {"availableTimeSec":27600,"steps":[{"name":"Cut","manualSec":6,"autoSec":32,"toolChangeParts":500,"toolChangeTime":2,"toolChangeUnit":"min"},{"name":"Fine Grind","manualSec":7,"autoSec":30,"toolChangeParts":200,"toolChangeTime":5}]}. Ask for missing available time or step times; never invent measurements or assume a shift length. Completion defaults to manual + auto; ask for the measured completion time when manual and auto overlap. Confirm the tool change unit when unclear. Do not claim the line meets demand; capacity excludes breakdowns, scrap, and demand. On corrections reuse the full previous input and change only the requested value. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
process_capacity_sheet
Use this when someone wants a SIPOC or a high-level process scope showing suppliers, inputs, process steps, outputs, and customers. Render five real columns from the supplied information. Keep unknown columns empty and ask one question about the most useful missing information. Do not invent suppliers, customers, steps, outputs, or scope boundaries. Partial maps are valid. Example: {"title":"Order fulfillment","process":["Pick","Pack","Ship"],"inputs":["Order","Stock"]}. Preserve process order. On a correction, reuse the entire previous input and change only the requested part. Use this standalone tool unless an A3 is requested. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
sipoc_diagram
Use this when someone wants to size the gain from a SMED or setup reduction: minutes saved per changeover, percent reduction, weekly minutes, annual capacity hours, and optional annual value. Example: {"processName":"Press Brake Tooling Change","currentStoppedMinutes":55,"targetStoppedMinutes":18,"changeoversPerWeek":9,"hourlyValue":185}. Ask for missing current time, target time, or changeovers per week; do not invent them. Omit hourlyValue when unknown; never guess a rate. weeksPerYear defaults to 50 as on the site; say so. The target cannot exceed the current time. This tool does not classify setup steps as internal or external; do not invent classifications. On corrections reuse the full previous input and change only the requested value. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
smed_analysis
Use this when someone wants a standardized work combination table (SWCT), or wants to check whether one operator's sequence of manual work, machine (auto) time, and walking fits takt time. Shows each step's start and end times, cycle time (manual + walk + wait), how far under or over takt, operator waiting, and machine time that runs past takt. Example: {"taktSeconds":45,"steps":[{"name":"Load press","manualSeconds":2},{"name":"Machine cycle","manualSeconds":1,"autoSeconds":6,"operatorDuringAuto":"waits"},{"name":"Apply label","manualSeconds":3,"walkSeconds":3}]}. Do not invent steps, times, or takt. Ask for any missing manual time or takt, or calculate takt with takt_time. Record walk time on the step it follows; the last step's walk is the return walk to step 1. Auto time counts toward the cycle only when the operator waits for the machine. On corrections reuse the full previous input and change only the requested values. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
standardized_work_combination_table
Use this when someone wants takt time, the pace a process must run to meet customer demand. Input demand per day and shift pattern. Returns an editable calculation in seconds and minutes. Verify the returned shift/break defaults with the user. Edit inputs directly in the widget or through chat. For updates, reuse the latest complete input (including documentId), preferring the manually edited draft, and change only requested values. Start a new document without documentId. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
takt_time
Use this when someone has timed work elements over several observed cycles and wants a time observation sheet, the lowest repeatable time per element, cycle totals, or which readings vary or look like outliers. Render the sheet immediately from the supplied readings. Example: {"process":"Final assembly","elements":[{"name":"Load fixture","readingsSec":[2.1,2.3,1.4,2.1]},{"name":"Inspect","readingsSec":[8.3,11,8.3,19.5]}]}. Readings are seconds in cycle order, up to 10 cycles; use null for a cycle that was not timed. Do not invent, average, round, or fill in readings, machine times, or names. The tool takes the lowest reading, to 0.1 s, that occurred at least twice, never the average. An element with no repeated reading has no repeatable time; say so and do not choose one yourself. Set selectedSec only when the user picks a different observed reading. Readings over 2× the selected time are flagged as outliers; ask what happened in those cycles rather than dropping them. On corrections or added cycles, reuse the full previous input and change only the requested values. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
time_observation
Use this when someone asks for a value stream map, or describes a process from supplier to customer with its steps and times, even when some waits or inventory are unknown. Draw an editable map with ordered steps and inventory in working days. Returns lead time, processing time, process time share (not confirmed value-added time), takt, and the slowest step. Draw the map right away from what the user gave, then ask one question about the most useful missing number; never hold the map back to ask first. Leave optional changeover, uptime, operators and parties unknown unless supplied. A step time the user has not given is null. inventoryDays needs one entry before each step and one after the last: use null for any inventory or wait the user has not measured, so the map shows it as unknown; never guess and never use 0 for unknown. For batch work, record the time for the whole batch on the step and say so in its name (for example 'Shape (100-loaf batch)'). A long wait inside the process (for example 10 hours of fermenting or curing) can be its own step with that time. Do not invent information-control routes. Edit inputs directly in the widget or through chat. For updates, reuse the latest complete input (including documentId), preferring the manually edited draft, and change only requested values. Start a new document without documentId. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
value_stream_map
Use this when someone describes how to do a job, wants a work instruction, job breakdown, SOP, or TWI Job Instruction sheet. Turn their explanation into a visual sheet of important steps (what), key points (how), and paired reasons (why). Render a partial sheet immediately from supplied details, then ask one focused question about a missing key point or reason. Example: {"title":"Pack a kit","steps":[{"importantStep":"Match the order","keyPoints":[{"text":"Read the SKU, not the color","type":"quality","reason":"Different products use the same color"}]}]}. Do not invent steps, reasons, settings, measurements, safety rules, names, dates, or revisions. Preserve exact supplied quantities and conditions. Each reason must stay attached to its own key point. Classify supplied points as safety, quality, or technique only when clear; otherwise unspecified. Unknown keyPoints can be empty and unknown reasons stay empty. Photos are not supported in this version; do not fabricate image URLs. For corrections, additions, removals, or reordering, reuse the entire returned input snapshot, retain stable step IDs and unchanged metadata/points/reasons, and change only what was requested. Omit id only for new steps. This standalone sheet supports direct editing and chat updates; do not call a3_report unless the user requests an A3. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
work_instruction
Use this when someone wants a Yamazumi chart, operator balance chart, or line balancing check: work content per operator compared with takt time. Supply takt in seconds and each operator's measured work elements in seconds, optionally split into value-added (va), necessary non-value-added (nnva), and waste (nva) seconds that add up to the element time. Returns per-operator totals against takt, the VA/NNVA/waste split, line balance rate, theoretical minimum operators with and without waste, and the bottleneck. Example: {"taktSec":60,"operators":[{"name":"OP-10","elements":[{"name":"Press bearing","seconds":15,"va":11,"nnva":1.5,"nva":2.5},{"name":"Walk to bin","seconds":6,"va":0,"nnva":0,"nva":6}]},{"name":"OP-20","elements":[{"name":"Install shaft","seconds":13}]}]}. Ask for a missing takt or element time; never invent times, splits, or operators. If only available time and demand are known, use takt_time first. Leave the split out for elements the user has not classified. On corrections, reuse the full previous input and change only the requested values. Do not call a3_report unless the user wants an A3. Users can edit directly in the widget. On continuation prefer the latest edited input, reuse its documentId, and preserve unchanged data. An applied edit is a complete replacement input, including removals. Omit documentId for a new document. Call this tool once per reply. Do not call it again in the same reply to rename, reword, or tidy the result; if something needs changing, say so and let the user decide.
yamazumi_chart
Recommended guideHow to get your ChatGPT Plugin suggestedFollow the practical steps agents use to find, shortlist, and choose your Plugin.Read guideKaizumi Lean Tools ChatGPT Plugin FAQ
How the directory, categories and Discoverability Score work.
Read the methodologyHow do I improve Kaizumi Lean Tools's ChatGPT Plugin discoverability?
The levers are the listing surface agents actually read: names, descriptions, keywords, tool metadata, and registry health. Which lever matters depends on where discovery breaks, which is what continuous measurement shows.
What are Kaizumi Lean Tools alternatives on ChatGPT?
As of 2026-10-09, Kaizumi Lean Tools competes with 101, AllEars.Vet, allmanager, Axle Connect, B2 Portal, Baukosten-Cockpit, BauLedger, Btoo and 51 more in ChatGPT ERP & Operations Resource Tools, ranked by public Discoverability Score.
Where is this profile measured?
This profile uses the geography attached to the latest public registry snapshot: US. Locale tags are intentionally omitted.