GeDenTo.AI
AI meets technical documentation for next-generation engineering.
GeDenTo.AI reads ISO, ASME, CEN, GB/T and your own internal standards, then answers an engineering question with a verdict — feasible, not feasible or depends — and the exact clause, page and figure it came from.
The platform is live, and a scoped trial runs against your own drawings rather than a demo dataset.
Standards interpreted today
- ISO 1101
- ISO 286
- ASME Y14.5
- CEN
- GB/T
- Your internal design rules
Answers returned in English, Swedish or Chinese, cited to the source in its original language.
You digitised the documents. Nobody digitised the judgement.
CAD, PLM and SRM store what was decided. They do not help the engineer who is standing in front of an ambiguous tolerance callout at 16:40 on a Friday, waiting on the one person in the building who knows the answer.
of manufacturing revenue lost to quality-related cost.
ASQ, Cost of Poor Quality benchmarkof a knowledge worker’s week goes to searching for information.
McKinsey Global Institutepeople per organisation hold the practical standards knowledge. It leaves when they do.
GeDenTo customer discoveryof quality engineers’ time spent interpreting standards rather than resolving quality issues.
GeDenTo customer discoveryEvidence first. Reasoning second. Never the other way round.
The single question every engineering and quality leader asks is “how do I know it isn’t making this up?”. This is the answer, in five steps.
-
01
Ingestion
Public standards, your internal design rules, drawings and scanned PDFs are converted into a machine-interpretable knowledge graph.
-
02
Retrieval before reasoning
The model is not permitted to generate evidence. It must retrieve it first, and every fragment carries document, section and page metadata.
-
03
Reasoning in engineering language
Standards logic becomes actionable guidance and worked examples, in the vocabulary a design or quality engineer already uses.
-
04
Validation layer
A separate agent cross-checks the reasoning against the retrieved evidence before anything is shown, and flags where the standard genuinely leaves room for interpretation.
-
05
Traceable delivery
A verdict, the conditions attached to it, and citations down to clause, page and figure — in a form that can sit in a design review or an audit trail.
Same decision layer, four different Mondays.
Design & production engineersThe everyday users
Resolve a tolerance or datum question in seconds instead of queueing for a specialist, and carry the citation straight into the design review.
Domain specialistsToday’s bottleneck
Your interpretation, made available to every engineer and every site, in every language, without you being copied into the thread.
Quality & supplier qualityCost of quality owners
Settle a conformance call with an auditable, standards-grounded trail — and cut the clarification cycles and NCRs that inconsistent interpretation produces.
Digital manufacturing & ITThe people who have to defend it
A GenAI deployment that survives a board review: evidence-grounded, auditable, inside your own tenant, with no vendor visibility into your data.
Model the value against your own headcount.
Built on GeDenTo’s documented value model. Six of its eight assumptions were checked against independent benchmarks and came back conservative or aligned; the remaining two are flagged for validation in your pilot.
Licence cost is modelled as a blended annual cost per user.
Get this modelled for your siteIt runs in your tenant. We see nothing.
The most common reason a GenAI project stalls in an industrial organisation is that legal and IT cannot answer where the data goes. Here, the answer is short.
- Your cloud or on-premiseDeployed inside your own tenant, on the infrastructure you already run.
- You own input and outputEvery question asked and every answer produced belongs to you. GeDenTo has zero visibility into it.
- Traceable by constructionCitations are not generated after the fact. They come from the retrieval metadata, so an answer without evidence cannot be produced.
- Access control and versioningStandards are versioned and governed, so an answer reflects the revision that actually applies to the programme.
- Multilingual by designAsk in one language, search standards written in another, get the answer in the language the recipient needs.
Built by the people who used to be the bottleneck.
“Standards-driven decisions are locked in the heads of a handful of senior engineers. We are turning them into a traceable decision layer the whole organisation can use.”
Mehdi Ghiassi
Co-founderFifteen years interpreting GD&T and standards by hand inside Scania, Volvo Cars, Volvo Trucks and Polestar. The product reflects what actually goes wrong on the floor.
Mirza Dzonlic
Co-founderBuilds and takes enterprise platforms to market in regulated industrial environments, where traceability and data ownership decide whether a system gets deployed at all.
Based in Stockholm. The platform is live and in active discovery with Nordic OEMs and the Swedish standards ecosystem.
The hard ones, answered directly.
How do I know the answer is not invented?
The reasoning model is never allowed to generate evidence — it has to retrieve it from ingested source material first, and every retrieved fragment carries document, section and page metadata. A separate validation agent then cross-checks the reasoning against that evidence. Traceability is a property of the architecture, not a feature layered on top.
Where does it run, and who owns the data?
Inside your own cloud tenant or on-premise. You own all input and output data, and GeDenTo has zero visibility into it. This is usually the question that decides whether an industrial GenAI project happens at all, so it is worth settling early.
We have our own internal standards, not just ISO or ASME.
That is exactly what it is built for. Your internal design rules, drawings and scanned PDFs are ingested alongside public standards, and new domains can be added without rebuilding the architecture.
How is this different from a chatbot over our documents?
A chatbot returns a summary and leaves the decision with you. GeDenTo returns the decision — feasible, not feasible or depends — together with the conditions attached to it and the clause it rests on. That difference is what lets it be used inside a design review or an audit trail rather than beside one.
How long until we see value?
Technical deployment typically takes about a week. Because the trial is scoped against a problem your champion already has rather than a canned demo, evidence starts accumulating in week one.
What if the ROI model does not fully hold?
Assume half of it. Even at 50% of modelled value, the documented model still returns roughly four times the licence cost within the year. The calculator above has a conservative toggle so you can see that case directly.
Does it work across languages and across standards?
Yes. A question asked in one language can search standards written in another — an ISO requirement and its GB/T counterpart, for example — and return the answer in the language the recipient needs, with both sources cited.
Bring us one question you keep arguing about.
Thirty minutes, no deck. We walk through where standards interpretation is actually costing your team time, and whether a scoped trial is worth running. If it is not, we will say so.
Prefer email? info@gedento.ai