Omicron AI Labs · Paris · AI studio

We build agents that decide, not dashboards that report.

Omicron AI Labs builds its own agentic products — and deploys the same architectures for its clients. PostIQ is the proof: it scores every post against that creator's own recent history, never a platform average, and hands back an action already drafted while the window is still open.

Explore PostIQ Our consulting work
Data hosted in Paris EU AI Act art. 12 traceability Human approval required

Flagship product

PostIQ — the intelligence layer for creator agencies

Instagram Insights, TikTok Analytics, Looker: every one of them tells you what already happened. PostIQ watches the first two hours of every post, catches the breakout and the drop alike against the account's own norm, and hands the team the action to take, already drafted.

See the product and pricing
  1. T+30

    First reading. The monitoring agent notices the post on its own — nobody has to remember to check.

  2. T+60

    Second reading. The gap either holds or fades against the account's own last 10 posts.

  3. T+90

    Third reading. The alert lands while cross-posting or boosting can still change the outcome.

  4. T+120

    Final reading. The recommendation is filed with its figures, its model and its prompt version.

The same post, measured three times

PostIQ card at the T+30 reading: 97 views, 8.8% retention, 0% gap against the account average, velocity curve showing a single point.
T+30One reading. Nothing to report: the gap against the account average is zero.
PostIQ card at the T+90 reading: 208 views, +40% against the account average, velocity curve climbing from T+30 to T+90.
T+90Three readings. Views hit +40%: the gap holds, and the alert goes out.
PostIQ card at the T+120 reading: 236 views, +38% against the account average, complete velocity curve from T+30 to T+120.
T+120Four readings. 236 views, +38%: the curve is complete and the recommendation filed.

The same screen, minutes apart

PostIQ dashboard still collecting data: "Not enough data yet to call a trend".
Baseline formingBefore the first three readings, PostIQ makes no call — it has nothing to compare against yet.
PostIQ dashboard flagging a trend: "This post is taking off", 2.1x the account baseline, +1.91 sigma.
Baseline set2.1× its own baseline, +1.91 σ — and an action to approve, while the window is still open.

Their history, not the platform's

Welford rolling statistics over each creator's own last 10 posts. 35,000 views can be routine on one account and exceptional on another — only one of them warrants an alert.

Five agents, one decision-maker

Monitoring, analytics, insight, publishing, record-keeping. The agents do the measuring, the maths and the writing. Your team keeps the deciding.

It drafts. You approve.

Nothing leaves the product without an authenticated person approving it. No automatic posting, on any connected account.

The receipts are kept

Every recommendation is stored with its figures, its reference history, its model and its prompt version — the record the EU AI Act's Article 12 describes, readable by a legal team.

How we build

What running our own product taught us

PostIQ runs in production, with real accounts, real regulatory constraints and real inference costs. The same principles shape the work we do for clients.

  • 01 A baseline before a threshold. A raw number means nothing until you know what to compare it to.
  • 02 A human on the critical path. The agent prepares and proposes; a named person commits.
  • 03 Traceability written in at design time, not bolted on before the audit.
  • 04 Production as the point where the work starts, not the finish line.

Consulting

The same architectures, inside your operations

An independent practice with no captive vendor partnership. We work with mid-market and enterprise teams from strategic framing through to production — every engagement scoped with explicit KPIs and measurable milestones.

See the consulting offer
  • AI strategy Maturity audit, high-ROI use cases, a roadmap prioritised over 6 to 18 months.
  • Agentic engineering, RAG & cloud Autonomous agents, RAG and GraphRAG systems, NLP pipelines and industrialised LLM integrations.
  • Training & enablement Bringing technical and business teams up to speed on GenAI, with lasting skills transfer.

Founder

Elhadji Ngom

Founder & Chief AI Architect. Eight years designing and deploying critical AI systems in enterprise environments, across Paris, Luxembourg and New York. He writes the code behind PostIQ and leads the consulting engagements.

View LinkedIn profile
  • GenAI in productionBedrock agents, AgentCore, Strands, LangGraph, MCP and A2A.
  • MLOps & cloudAWS and Azure, Terraform, CI/CD, monitoring and continuous evaluation.
  • ComplianceEU AI Act, GDPR, audit trail and data sovereignty.