AI COMPMATE, THE THINKING LAYER IN ACTIVE DEVELOPMENT / 2026
§ Organisational AI OS · enterprise client In active development for first enterprise client

The company that
thinks about itself.

AI CompMate is an organisational AI operating system, a thinking layer inside Microsoft Teams that reads how the business actually works, sets its own objectives, designs the plans to hit them, builds the agents that run them, and grades its own results. Autonomous in thinking, governed in acting. It is in active development for its first enterprise deployment; the numbers below are design targets.

Built on the standards that shipped TAFI 279/279, Orions 55/55 and VistaGPT 95.3/100.

0 + hrs recovered per employee / week, target
~0 % of attrition caught early, target
0 % less manual reporting, target
~$300 /mo design target vs a $240K/yr leadership layer
REASONING CORE Objective Blueprint Approval ✓ Execute Adjust 100+ SIGNALS pulled passively · 24/7 GOVERNED ACTION learns & adjusts ↻
FIG.00, Signal in, governed action out, learning back · illustrative, in development
§ Inside Teams

The CEO morning brief,
delivered as a message.

The thinking layer lives where work already happens. This is the intended shape of the daily brief and an assistant reply, rendered here as a design mock, not a live product screenshot.

FIG.01, CEO morning brief & assistant reply in Microsoft Teams · illustrative, in development
Representative format only. AI CompMate is in active development for its first enterprise client, this is not a captured product screenshot.
§01, The problem

Your company has
a thinking problem.

Every growing company has a leadership layer it can’t fully staff, the Chief of Staff, the COO, the analyst for every manager, the assistant for every employee. The roles that should exist, but can’t be afforded. That layer costs upward of $240K a year.

And it still misses things. The signals that matter, a slipping deal, a flight-risk employee, a stalled decision, live in message threads and meeting transcripts, not the monthly reports humans read.

Roughly 50,000 words of signal pass through Teams every day. Almost all of it evaporates. AI CompMate is being built to read it, and then think about what to do.

$240K a year for a leadership layer that still can’t read everything, all the time
~$300 a month is the design target for a thinking layer that does, running 24/7, inside Teams (₹2Cr/yr → ~$300/month)
§02, The autonomous loop

Observe, propose, design,
execute, measure, adjust.

The part that makes it more than an assistant. The loop never stops, and once you lock a goal, it owns the follow-through.

01Observe 02Identify 03Propose 04Design 05Assemble 06Execute 07Measure 08Adjust THE LOOP never stops
Autonomous in thinking, it decides what’s worth doing, and how Governed in acting, activating a goal or any external write needs a human yes
01Observe

9 passive intelligence streams

02Identify

gaps & opportunities

03Propose

objectives + evidence

04Design

a dependency graph of steps

05Assemble

ephemeral agent team

06Execute

approval-gated writes

07Measure

impact vs forecast

08Adjust

re-checked every 14 days

§03, How it thinks for itself

Six capabilities most
“AI agents” don’t have.

The difference between a tool you prompt and a system that thinks: it decides what’s worth doing, does it, checks whether it worked, and improves, inside a governed frame.

It sets its own objectives

Gap and opportunity scans run against the live model of the business. Each objective is scored on impact × urgency × feasibility.

capped to 5–7 active at once

It designs its own blueprint

A planner turns a locked objective into a validated graph, dependencies, parallel branches, approval checkpoints, a cost estimate and a fallback.

prefers the simplest fix

It builds its own agents

Agents don’t exist until a task demands them. A factory writes the prompt, selects the tools, scopes the knowledge, then the agent runs and dissolves.

only the learnings persist

It checks itself against its metrics

Every objective and agent declares success criteria up front. Each run is measured against them and rolled into green / amber / red performance cards.

maths, not vibes

It learns and gets smarter

Winning assemblies are promoted into a pattern library; failures recorded as anti-patterns. The next similar task reuses the proven pattern, faster, cheaper, better.

a compounding flywheel

It locks on and adjusts in real time

Once you approve a goal, a monitor tracks every step daily, nudging owners on day 2, escalating by day 5, re-checking relevance every 14 days. It stays on the goal.

real-time follow-through
§ Reasoning

It traces cause
and effect.

A live causal graph turns scattered symptoms into a root-cause chain. Not “delivery is red”, the chain of decisions that made it red, every edge confidence-scored.

FIG.03, A root-cause chain, traced through the causal graph
0.78 0.85 0.81 0.64 Unclear PRD requirements gap Rework +18% scope churn Velocity −23% vs 4-week avg Delivery: RED sprint will slip Client: at risk 2 anxious emails
Root cause: unclear requirements on Sprint 5, chain confidence 0.55 (product of edges). Fix the cause, not the symptom.
§ Use cases in action

See it think,
case by case.

Every scenario runs the same three beats, read the signal, reason to a conclusion, propose the action for your approval. Six on rotation; click any to hold.

She hasn’t said she’s leaving. The signals already have.

Read Behavioural shift Message volume down 60%, tone flat, two 1:1s skipped.
Reason Flight-risk 0.78 Three weeks of decline, the pattern matches past attrition.
Act Retention nudge A private heads-up and talking points drafted for her manager. Approve ✓

Your best candidate is going quiet, and so is your shot at hiring them.

Read 3 candidates idle Top of the pipeline untouched for 14+ days.
Reason Going cold The best-fit hire is likely fielding a competing offer.
Act Follow-ups ready Personalised emails and interview slots proposed. Approve ✓

The account looks fine on the invoice. The meetings say otherwise.

Read Tone cooling Client asking about timelines, mentioning other vendors.
Reason Health 2 / 5 Renewal at risk within 60 days.
Act Exec check-in A meeting booked and recovery talking points drafted. Approve ✓

Someone promised. Nobody followed up, until now.

Read Promise made “I’ll ship the API by Wed.” Wednesday passed, no commit.
Reason 24h overdue Blocks two downstream tasks; no follow-up detected.
Act Gentle nudge The owner reminded, the item flagged for standup. Approve ✓

The meeting ended. The minutes wrote themselves.

Read Call ended A 45-minute Teams meeting, full transcript captured.
Reason Extracted 5 decisions, 3 action items, 1 blocker, each with an owner.
Act Minutes sent Summary emailed to attendees, tasks created. Approve ✓

212 messages, 9 meetings. One thing actually matters today.

Read Last 24 hours Every message, meeting and alert across the company.
Reason The one thing Ranked by impact × urgency, plus five things to ignore.
Act Brief delivered A single prioritised message, waiting in Teams at 7:42. Delivered
§ Proof, the loop, running

We fed it a 3-month chat. It caught a leaked password.

We handed the system a raw Teams chat history with no instruction beyond “make sense of this.” On its own it comprehended the intent, identified five domains, resolved the people and systems, and generated a five-step plan.

It assembled five ephemeral agents, ran them, and produced a structured brief, tasks with owners, open blockers, and a P0 security incident. Buried in the thread, an engineer had pasted a plaintext service-account credential. It scored it CVSS 9.8 and wrote a full remediation runbook. No one told it to look for security issues. It reasoned that it mattered, and led with it.

FIG.02, Autonomous run, 5 agents assembled on the fly
STEP 1 Parser normalise + resolve entities
STEP 2 Task extractor owners, deps, due dates
STEP 3 Blocker analyst open questions
STEP 4 Security auditor CVSS 9.8 flagged
STEP 5 Synthesiser one prioritised brief
§04, The self-building framework

Eight primitives.
The platform builds itself.

Every operable concept is a versioned, frozen, auditable row, not code. A new agent is a new row; a new system, a connector. AI-enable a system in days of metadata, not weeks of code.

PromptBlueprint

Versioned, templated LLM instructions

ToolBlueprint

A capability contract, inputs, outputs, side-effects, approval flag

SkillBlueprint

A named composition of tools with explicit I/O and rollback

AgentBlueprint

Persona + prompt + tools + skills + knowledge + policy

WorkflowBlueprint

Triggered sequences, cron, event, or manual

ConnectorBlueprint

Catalog-first definition of an external system

ObjectiveBlueprint

What the system is trying to achieve, with benchmarks

KnowledgePackBlueprint

Bounded RAG scope, store, filter, permissions

§05, Cutting the noise

Nine passive streams.
Zero extra effort.

You already generate the signal, 200+ messages a day, 5–10 meetings. The system extracts ~65% of what it needs from work that’s already happening.

Decisions made

A decision log, built automatically, who decided what, when, reversible or not.

Commitments & follow-through

Every promise captured, then tracked to completion or escalation.

Blockers & friction

Repeated asks and escalation language surfaced in real time, not at the retro.

Sentiment & morale

Shifts in tone, volume and participation become flight-risk evidence.

Client health

Concern and warmth read from client meetings and internal chatter.

Time allocation

Calendar and presence reveal where attention goes versus where it should.

Attention patterns

All signal clustered by topic, “65% on firefighting, 5% on strategy.”

Network map

Who talks to whom, hubs, bridges, isolated nodes, key-man risk.

Pattern discovery

“Developers who attend 3+ trainings have 40% lower attrition.”

§06, The advanced AI behind it

The techniques
doing the work.

A causal graph, a planner, an evaluation harness, a temporal memory model and a multi-layer safety pipeline, working together.

ORCH
Dynamic multi-agent orchestration

A LangGraph supervisor + CrewAI crews route intent and dispatch ephemeral teams generated per task, not pre-wired.

CAUSAL
Causal-graph reasoning

A Neo4j graph models cause→effect with confidence-scored chains, why delivery slipped, not just that it did.

DAG
DAG planning & live replanning

A request compiles into a validated execution graph; a failed step triggers a replan, not a restart.

MEM
Bitemporal memory

Temporal facts answer “what was true at time T”, reasoning over how state evolved, not just today.

RRF
Hybrid retrieval (RRF)

Qdrant fuses dense + sparse search and LLM-reranks, precise recall across organisational knowledge.

JUDGE
LLM-as-judge evaluation

Outputs scored on faithfulness, relevancy and goal accuracy against golden sets, with A/B prompt tests.

ROUTE
Tiered model routing

A LiteLLM gateway routes by complexity, Haiku to classify, Sonnet to analyse, Opus to reason.

DURABLE
Durable, resumable workflows

Temporal wraps every chain, a crash at step 4 of 7 resumes at step 4. Autonomous work survives failure.

PRIVACY
Privacy-by-design ingestion

Restricted content is dropped before any model sees it; PII is detected and encrypted at rest.

QUAR
Quarantine over guessing

Ambiguous entities are held for review, never guessed. A wrong attribution can corrupt the graph for weeks.

LEARN
Pattern-library learning

A system-level memory of what worked and what failed feeds every future assembly decision.

META
Autonomy-by-metadata

Every operable concept is a versioned, frozen, auditable row, not code. The platform extends itself, under approval.

§07, What it does for you

From cutting the noise
to running the loop.

One system, three depths of help, earned in that order. It starts by giving everyone their time back, and grows into the executive layer you could never afford.

Use case 01

Cut the noise

It reads everything so your team doesn’t have to. 50,000 words of daily Teams signal become decisions, commitments and blockers, surfaced, not buried.

Every employee
Use case 02

A working layer in Teams

A scoped AI that preps your meetings, drafts timesheets from observed work, chases the commitments made to you, and answers questions about the business with cited evidence.

Every employee
Use case 03

An autonomous partner

It sets goals, plans them, runs the agents, measures impact and adjusts, you hold the approvals. The leadership layer you could never staff, running 24/7 under governance.

Leadership
§08, The agent workforce

33 specialist agents,
in 7 tiers.

One agent, one domain. Signal flows up the tiers, from raw ingestion through autonomous thinking to governed execution.

T1 Ingestion Meeting, chat, calendar and presence agents turn raw signal into structured intelligence. 4 agents
T2 Resolution Deduplicate commitments, detect contradicted decisions, link intelligence across sources. 3 agents
T3 Analysis Follow-through, client health, people trajectory, blockers, and the Machine Diagnostician. 7 agents
T4 Deep Analytics Behavioural profiles, problem-pattern engines, benchmarks, environmental and goal tracking. 5 agents
T5 Autonomous Lifecycle The thinking layer, objective generator, activity designer, execution monitor, impact & relevance assessors. 5 agents
T6 Delivery CEO brief, daily priority, weekly report, blocker alerts, anti-pattern and minutes delivery. 6 agents
T7 Executor Notifications, reminders and scheduling, every external action approval-gated. 3 agents
§09, Governed autonomy

Trust is earned,
not demanded.

Autonomy isn’t a switch you flip, it’s a ladder the system climbs by proving accuracy. Every external write is gated; one-way doors never auto-execute.

Stage 1

Observer

Month 1–2

Reads everything. Surfaces insight. Humans do all the thinking.

Gate: observations accurate >80%

Stage 2

Advisor

Month 3–4

Diagnoses root causes and patterns. Drafts for human review.

Gate: diagnoses validated >70%

Stage 3

Assistant

Month 5–6

Proposes objectives, designs activities. Approval card for every write.

Gate: recommendations succeed >60%

Stage 4

Partner

Month 8+

Auto-executes reversible work within boundaries you set.

One-way doors always need a human

§ Target outcomes by role

What it is built to deliver,
role by role.

Every target is tied to a mechanism, to be measured against real data during the first enterprise deployment. These are design targets, not achieved results.

Every employee 5+ hrs/week recovered Daily priority brief + commitment tracking
Every employee 80% less manual reporting Passive intelligence, no new input
HR 3–4 hrs/week recovered by Week 10 HR Pipeline Health agent + scheduler
HR ~60% of attrition caught early Flight-risk passive stream
CEO $240K/yr → ~$300/month Attention Allocator + agent workforce
Delivery 30% better decision quality Machine Diagnostician + causal chains
§ Where it’s going

A product, a framework,
then an advisor.

The same 8 primitives that let AI CompMate build itself are designed to outgrow it.

STAGE A Today

The thinking layer

The working AI operating system inside Microsoft Teams, the autonomous loop, the specialist agents, approval gates, measurable quality, bounded cost. First enterprise client: an enterprise software company.

STAGE B Next

Embeddable framework

The 8 primitives become a library and runtime another team can drop into their own product, chat-first AI against their own data and tools, with governance built in.

STAGE C Horizon

Auto-discovery advisor

Point it at an existing system. It identifies which workflows are AI-amenable and proposes the blueprint shapes. The framework helping build the framework. Humans still ship.

§ The stack

Built to be owned.
Plugged into your tools.

100% open-source and self-hostable, the only paid dependency is the LLM. Agents reach external systems only through a secured, audited boundary.

Orchestration
LangGraphCrewAIPydantic AITemporalFastMCPNATS JetStream
Memory & Knowledge
Postgres + pgvectorQdrantNeo4jMem0Zep GraphitiRedis
Routing & Evaluation
LiteLLMClaude (Haiku/Sonnet/Opus)LangfuseRagas
Platform
KeycloakInfisicalK3sArgo CDLGTM stack
Connects to
Microsoft TeamsOutlook EmailCalendarAirtable (ATS)Keka (HRMS)UpworkElevateHQLinkedInSharePointWhatsAppJira+ anything via MCP
Related work

Want a thinking layer
for your company?

AI CompMate is in active development for its first enterprise client. Tell us how your team works today, and we’ll show you what it would read, what it would propose, and where you’d hold the yes.

AGENTIX TECH · agentixtech.ai, architecture-first AI engineering