Kenop v1.3 · now available

Group brainstorming. Math first. Many LLMs.

Five capabilities, one platform. A team workspace where your engineers, finance, and ops solve real problems — with first-principles math doing the heavy lifting and the LLM of your choice on top.

Core principle

Math first. Prediction next.

Generic AI chatbots hallucinate numbers because they predict text. Kenop calculates from first principles — then asks the model to reason on top of an exact answer.

Step 1
Your data
Plant data · ERP · batch sheet · problem brief
Step 2 · The moat
Math runs first
SAP · stoichiometry · mass balance — exact, deterministic
Step 3
LLM reasons on top
Models interpret, compare, write the report
Feature 01

Consulting-grade investigation on complex problems.

The kind of structured root cause analysis that used to require a specialist consultant on-site for 3-5 days — available in under 15 minutes. Submit your problem, attach your data, and receive a report benchmarked against industry standards.

A 12-stage pipeline applies first-principles reasoning, causal chain analysis, protocol benchmarking, and adversarial review. The QA stage checks every recommendation against a numeric target before delivery.

Root cause analysis ranked by Bayesian likelihood %
Method verification per identified cause
Parameter status table — PASS / WARN / FAIL against benchmarks
Prioritised recommendations with numeric targets
Immediate 24-hour action checklist with assigned roles
Kenop Intelligence
Khandwa Oil Refinery — Investigation
Running
Trend scanClaude Opus
DecomposeClaude Opus
First principlesClaude Opus
Causal chainsGPT-4o
Protocol checkPerplexity
06
Deep researchPerplexity
07
Industry pulseGrok-3
08
SynthesisGPT-4o
09
Counter-argumentClaude Opus
10
Write reportClaude Sonnet
11
QA reviewClaude Sonnet
12
Structure JSONClaude Haiku
Process investigation thread
Memory transferred across models
ClaudeGPT-4oGrok
NK
Why is our acid oil FFA stuck below 62% even after adjusting the NaOH dosage last week?
CLAUDE OPUS · remembers your last 3 sessions
Based on Tuesday's degumming data — phosphorus at 12 ppm — NaOH adjustment alone cannot resolve the FFA ceiling. Limiting factor is soap stock quality (NHP carryover). TFM 35.35% confirms it: soap stock has too much entrained neutral oil to yield high-FFA acid oil regardless of lye strength.
K
NK
Switch to GPT-4o and ask it to model expected FFA% if we bring phosphorus to 5 ppm.
GPT-4o · context from Claude carried forward
With phosphorus ≤5 ppm and current FFA 3.2%, theoretical soap stock FFA improves to 62-65%. Acid oil FFA should reach 67-70% — above your ≥65% target. Adjust H₃PO₄ to 0.18% w/w, 20-min contact at 75°C.
K
Session context preserved across models
Feature 02

Switch models mid-conversation. Memory carries.

Switch between Claude Opus, GPT-4o, Grok, and Perplexity mid-conversation — without losing context. Each model receives the full thread history and picks up where the previous one left off.

Use the analytical strength of one model to inform the reasoning of another — in a single coherent conversation about your data.

Persistent memory
Session context survives model switches. No re-explaining.
Model specialisation
Perplexity for literature, Claude for reasoning, GPT-4o for math.
Your data, always
Every model sees your uploaded files and prior analysis.
Full history
Every response, every switch, every finding — logged.
Feature 03

Kenop joins your team as an active member.

A shared workspace inside the Kenop app. Add your engineers, finance, plant operators. Everyone queries the same data, sees the same threads, builds on the same findings.

Kenop is one of the members. When a teammate flags a problem in the workspace, it applies global compute automatically — without waiting for a separate investigation request.

What this looks like in practice
Your QA manager flags a colour deviation in the bleached oil. Kenop sees it in the workspace, pulls the last 14 days of bleaching data, cross-references against historical readings, and posts a structured root cause analysis — before the process engineer has opened the file.
Inside the Kenop app · no third-party messenger required
Khandwa Oil — Team workspace
3 members active
NK
SR
AM
KI
Nachiket, Shradha, Abhay + Kenop Intelligence
Kenop applies compute when a problem is flagged
SR
Shradha09:14FLAGGED
Colour reading came back at 4.8R this morning — above limit. Third time this week. Something is wrong with bleaching.
KI
Kenop Intelligence09:15AI MEMBER
Flagged. Pulling bleaching data for the last 14 days. Running analysis against historical colour baseline and earth dosage records. Will post findings in 8 minutes.
NK
Nachiket09:16
Thank you. Check if earth dosage changed after the supplier switch on April 12.
KI
Kenop Intelligence09:23AI MEMBER
Analysis complete. Root cause: bleaching earth oil retention increased from 28% → 34% post supplier switch. New earth has higher bulk density, requires +0.25% w/w dosage. Recommendation: increase to 1.25% w/w, test on next batch.
Model selection
Pick your primary. Kenop routes each stage.
Each stage routes to the optimal model automatically
Feature 04

Your choice of model. Our routing.

Choose your primary model. Kenop intelligently routes each stage of your investigation to the model best suited to that specific task — regardless of your primary selection. Or override per stage if you have a preference.

When your private fine-tuned model is available, it participates in the pipeline for pattern-recognition tasks — while cloud models handle the broad reasoning. The compute is additive.

No lock-in
Switch primary models any time. Investigations are not tied to one provider.
Automatic fallback
If your selected model is unavailable, the next best picks up without interruption.
Private model integration
Your fine-tuned on-prem model participates alongside cloud models. On request.
Feature 05BETA

Your organisation's knowledge base.

Every investigation your team runs becomes part of an indexed organisational memory. When a new problem arises, Kenop searches your history first — finding prior root cause analyses, resolved anomalies, and successful interventions before running the full pipeline.

Your second colour-deviation investigation is faster and more accurate than the first — because Kenop already knows what worked, what parameters were involved, and what the engineer's corrective action achieved.

Beta access — on request
Storage memory is currently in beta. Available on request for organisations with three or more investigations on the platform.
Get accessRequest beta
Organisational knowledge base
BETA
3 prior investigations found
Colour deviation — Lovibond 4.8RApril 17, 2026
Root cause: Bleaching earth oil retention increased post supplier switch
Outcome: Dosage increased to 1.25% — resolved in 1 batch
BleachingColourEarth dosage
87% match
Repeated colour exceedance — Q1 2026January 22, 2026
Root cause: Residual soap in oil entering bleaching stage
Outcome: NaOH dosage reduced, soap removal improved
BleachingNeutralisationSoap
82% match
Bleaching earth performance dropNovember 8, 2025
Root cause: Moisture in incoming crude oil — earth saturation
Outcome: Pre-drying step added before bleaching
BleachingMoisturePre-treatment
79% match
Kenop v1.3

Everything above is available now. Log in to begin.

Five capabilities. One platform. Built for industrial operations that have outgrown generic AI tools.

Log in to Kenop →Try the investigation engine