Skip to content

AI consulting

Clients buy outcomes, not AI. We diagnose before we build, prove one use case on your own data, and keep every byte inside your walls.

sprint / day-3 / opportunity-matrix

Northline Freight · sample

Seven candidates. One gets built.

Day 3 of 7
Quick wins
Effort →↑ Impact

Chosen

Dispatch re-planning

Evals (real cases)

41 / 50 pass

9 failures shown

Conservative case

11 h / week

hours recovered

The gap

The opportunity is the gap between using AI and profiting from it.

Most firms already pay for AI. Far fewer can point to a line on the P&L that moved. The difference is rarely the model — it is choosing the right workflow, redesigning it, and measuring honestly.

  • of organisations use AI

    0%
  • see any EBIT impact from it

    0%
  • qualify as AI high performers

    0%

51 points between using AI and profiting from it. That is where we work.

Source: McKinsey, The State of AI, 2026

How engagements grow

Start small. Earn the next step.

Each rung pays for itself before the next one is proposed. You can stop at any of them and keep everything we built.

  1. 01

    Workshops

    Half-day to multi-day

    A shared, plain-language picture of what AI can and cannot do for your business — before anyone buys anything.

    What you get

    • Executive briefing on where value sits in your sector
    • Hands-on sessions for division and operations leads
    • A first-pass AI usage and data policy
  2. The paid diagnosis. We find the workflows where AI is worth the most and prove one on your own data.

    What you get

    • Opportunity map and AI data-exposure check
    • A working demo, scored on real cases
    • Conservative ROI, risk register and roadmap
  3. 03

    Quick-win agent build

    1–2 weeks

    One roadmap item, built around a redesigned workflow — not bolted onto the old one.

    What you get

    • One agent, one job, inside the tools you already use
    • Human approval on every high-risk action
    • Eval set and tool risk table handed over
  4. 04

    Care plan

    Monthly · cancel any month

    We stay after handover, because systems that are not watched quietly decay.

    What you get

    • Monitoring, error alerts and weekly health checks
    • Fixes within an agreed response time
    • One improvement a month, plus a cost and ROI report
  5. 05

    Private AI platform

    On your infrastructure

    Once two or three quick wins are earning, the Anoryx EcoSystem gives them a governed, self-hosted home.

    What you get

    • Ownership, spend, comms and orchestration in one stack
    • Deployed inside your environment — no data leaves
    • Honest status on every product — most are still in development

AI Opportunity Sprint

Proof on your own data in seven days.

A fixed-scope diagnosis that ends with a working demo, an honest eval and a number you can defend to your board — not a slide of possibilities.

You leave with

  • 3–5 stakeholder and staff interviews
  • Opportunity map across operations, sales and support
  • AI data-exposure check, including shadow AI
  • One working demo on your data, inside your control
  • Eval score on 20–50 real cases, failures shown
  • ROI model on conservative assumptions
  • Risk register structured on the NIST AI RMF
  • Roadmap: quick wins (months 1–3), medium (4–8), transformational (9–18)

Our bar before any readout

The demo ran on real dataFailures are visible in the evalEvery tool is risk-ratedConservative ROI comes firstA named owner after go-live
  1. Day 0

    Kickoff and charter

    We confirm the sponsor, the data owner and the one metric that matters, and sign a data-handling note before we touch anything.

    OutputOne-page charter

  2. Day 1

    Interviews

    Three to five conversations with leaders and front-line staff, 30–45 minutes each. We listen far more than we talk.

    OutputTime sinks and quality risks

  3. Day 2

    Map the work — and the shadow AI

    Operations, sales and support, mapped end to end — plus the AI tools your people already use without anyone signing off.

    OutputOpportunity canvas and data-exposure map

  4. Day 3

    Pick one use case

    Every candidate scored on impact and effort, run through the agent-or-automation test. We commit to one.

    OutputChosen use case and success metric

  5. Days 4–5

    A working demo on your data

    Built inside your environment and scored against 20–50 of your real cases. The failures go in the report, not under the rug.

    OutputWorking demo and eval score

  6. Day 6

    Size it honestly

    The conservative ROI case first, a risk register structured on the NIST AI RMF, and what it would cost to run each month.

    OutputROI sheet, risk register, cost estimate

  7. Day 7

    Readout and proposal

    Six slides for your sponsor and leadership, and a proposal for the build and care plan — or a plain ‘not yet’ if the numbers do not hold.

    OutputReadout and proposal

Identity Debt Assessment

Who — or what — can act inside your company?

A fixed-scope engagement, typically three to four weeks, that maps every human, machine and AI agent that can log in, spend money or read data — and who answers for each one.

Week 1

Read-only discovery

Identity provider, cloud, code, SaaS and the HR leaver list — inventoried with read-only access. Nothing in your environment changes.

Week 2

Ownership map

Every human, service account, key, AI agent and shadow app, linked to an accountable owner — or flagged as having none.

Week 3

Risk register and evidence gaps

What each actor can reach, ranked by blast radius, plus a gap list against DPDP, NIS2 and ISO 27001 evidence expectations.

Week 4

Remediation roadmap

What to fix first, who owns each fix, and what can be automated — so the map stays true after we leave.

assessment / ownership-map / register

Actors found

1,284

No owner

31%

Evidence gaps

17

read-only
ActorOwnerRisk

svc-payroll-export

Service account

unownedHigh

procurement-agent

AI agent

R. MenonMedium

ci-deploy-token

API token

unownedHigh

notes-ai (browser)

Shadow AI app

unownedHigh

s.kapoor (left Jun)

Former employee

HRMedium

backup-reader

Service account

IT OpsLow

The output is audit-ready evidence that maps to DPDP, NIS2 and ISO 27001 — not a certificate. When you want the map kept current, it carries over into Sentinel, self-hosted.

How we work

Eight rules we do not bend.

They make us slower to say yes and faster to be trusted.

Diagnose before we build

Prescription without diagnosis is malpractice. Every engagement starts with the work, not the model.

One agent, one job

Narrow agents are testable, explainable and easy to switch off. Generalists are none of those.

Humans approve high-risk actions

No agent moves money or signs commitments without a person saying yes.

Staged autonomy

Drafts first. Then one action. Then more — only after weeks of clean logs.

Show the failure cases

Honest evals win more trust than polished demos. You see what breaks, and how often.

Conservative ROI only

Hours, errors, money and risk, measured in your units. We lead with the cautious case.

Your data stays in your control

Data never leaves the client: architecture, not adjectives. Exceptions only in writing.

Inside the tools you already use

Nobody wants a tech stack on top of a tech stack. We deliver where your people already work.

Where we focus

Built for firms that own physical work.

Asset-heavy and regulated mid-market companies, where a dispatcher’s hour or an auditor’s question has a real price. This is our focus, not a client list.

Logistics

Fleet, warehousing and freight forwarding — dispatch, PODs, detention charges.

Construction

Quotes, RFIs, procurement and subcontractor paperwork.

Facilities

Service tickets, maintenance scheduling and quote follow-up.

Hospitality

Guest requests, rostering and supplier coordination.

Regulated mid-market

BFSI, health and IT services, where identity and evidence matter as much as speed.

Example use cases

Illustrative

Dispatch co-pilot

Re-plans routes when a driver calls in sick or a load slips, and drafts the customer update.

Dispatcher approves every change

Quote drafting

Turns an emailed request and site photos into a priced draft from your own rate card.

Estimator signs off before it is sent

Procurement agent

Compares supplier offers against spec and history, and prepares the purchase order.

Above a threshold, a human approves

Support triage

Reads incoming tickets, classifies and routes them, and drafts a first reply.

Drafts reviewed until the logs earn trust

Try it

Agent or automation?

Not everything needs an agent. Most things do not. Answer four questions about one piece of work and see what we would recommend.

Based on the agent-versus-automation test we run on Day 3 of every Sprint.

Question 1 of 4

Does the work arrive as structured data?

Form fields, spreadsheet rows, system records — rather than emails, PDFs, photos or voice notes.

A rule of thumb, not a substitute for looking at the work. In a Sprint we run this test on every candidate with the people who do the job.

Once quick wins are earning, the platform gives them a governed, self-hosted home. Not before.

Explore the EcoSystem

Find the workflow worth fixing. Prove it on your data in a week.

Tell us where time, money or risk is leaking. We will tell you plainly whether a Sprint makes sense — and what we would look at first.