Most service desk AI programmes we are asked to review started as a chatbot pilot. The chatbot answered the top twenty questions well, deflection climbed for a quarter, and then it fell back — not because the model degraded, but because the knowledge behind it did. Nobody owned the articles. Nobody reviewed them. The AI kept answering confidently from a knowledge base that no longer matched operational reality.
The instinct is to change the model, or change the vendor, or change the surface. The instinct is wrong. Deflection is not a model problem — it is an operating-model problem. B7 treats AI triage as an ITSM discipline: integrated into the platform your agents already work in, grounded in knowledge that has a named owner and a review SLA, measured on a dashboard your service desk manager reads weekly. That is what makes deflection sustain past the first quarter.
Five streams,
ending in sustained deflection.
Baseline and knowledge audit front-load in weeks 1–2. Triage build and KnowledgeOps overlap through weeks 3–7. Pilot measurement and handover close weeks 8–10.
Baseline & ticket taxonomy
Twelve months of ticket history analysed. Volume by category, first-contact resolution rate, aging distribution, and reopen rate captured as the pre-engagement baseline every later claim is measured against.
Knowledge audit & KnowledgeOps design
Existing articles assessed for currency, coverage, and answerability. Named owner assigned per knowledge domain. Review cadence and change audit trail agreed with the service desk manager.
ITSM-integrated triage build
Classification against your live service catalog, routing to existing resolution paths, RAG grounded on governed sources. Built inside your ITSM platform — no platform swap, no parallel tool for agents to learn.
Evaluation harness & escalation rules
Gold-standard ticket set curated with senior agents. Confidence thresholds set per category. Escalation to a human agent wherever the threshold is not met — a wrong answer costs more than a routed one.
Pilot measurement & handover
Live pilot on a scoped ticket population. Deflection dashboard reporting against baseline. Runbooks written for triage drift, knowledge staleness, and rollback. Handover to service desk operations.
Ten weeks maximum.
Six minimum. Four phases.
Phase count is fixed. Duration flexes with knowledge-base condition, ITSM integration complexity, and the number of ticket categories in pilot scope. Milestones are signed gates — not aspirations.
Ticket categories scored,
not chosen by complaint volume.
Every candidate ticket category runs a six-criteria scorecard in weeks 1–2. Each criterion scored 1–5 against ticket evidence, not agent opinion. Signed by the service desk manager before Phase 2 begins.
From an ungoverned chatbot
to governed triage.
A typical UAE service desk arrives with a portal chatbot bolted beside the ITSM platform, a knowledge base nobody owns, and deflection reported from the chatbot vendor's own console. The engagement moves the discipline inside the platform of record.
Reference pattern. Some organisations retain a portal chat surface as one entry point into the same governed triage layer. What always changes is that the deflection claim comes from the platform of record, measured against a baseline captured before the engagement began.
A service desk
under volume pressure.
Representative pattern for a UAE logistics group of this scale — 6,400 staff, 11,000 tickets per month, 22-agent service desk on Sunday–Thursday cover. Ranges reflect target outcomes NexITC underwrites in scope for this class of engagement. N=1 — illustrative composite, not a specific client.
Four artifacts,
each with signed acceptance.
Every deliverable has documented acceptance criteria signed at engagement kickoff. Nothing more, nothing less.
ITSM-Integrated AI Triage
Classification, routing, and RAG-grounded assist running inside your existing service management platform. Confidence thresholds and human escalation rules configured per category. Configuration-as-code handed over.
KnowledgeOps Governance Pack
Named owner per knowledge domain, review cadence and SLA, change audit trail, staleness detection workflow. The artifact that keeps deflection from decaying after quarter one.
Deflection Dashboard
Deflection rate, first-contact resolution, ticket aging (median and P90), and resolution time trend — all against the pre-engagement baseline, reported inside the platform of record.
Service Desk Operations Runbooks
Runbooks for triage drift, knowledge staleness, threshold tuning, and rollback — written for the service desk manager and rehearsed with the team before handover. This is the deliverable that decides whether the discipline survives without us.
Six outcome metrics,
measured pre and post.
Success is not “the AI triage is deployed.” It is measured against six specific outcomes captured in a baseline report at engagement start and re-measured at pilot steady state.
Honest scoping.
B7 is a fit when specific conditions are met. It is not a fit when other conditions are. We say so before the scope conversation, not after the commercial commitment.
Twelve months of ticket data with category and resolution fields populated. The baseline is derived from your data, not from benchmarks.
Signs off the category scorecard, confidence thresholds, and handover. Typically 25% time commitment across the engagement.
KnowledgeOps requires named owners inside your organisation. If nobody will own knowledge currency, deflection will decay whatever we build.
Read and write scope agreed with the platform administrator before Phase 2. Access delays are the most common cause of B7 timeline slippage.
Ticket bodies frequently carry personal data. Residency and retention posture confirmed before any content reaches a model.
B7 is an internal service desk discipline. Customer-facing assistants grounded on enterprise knowledge are B2 GenAI Knowledge Assistant Build™.
If MTTR, change governance, and infrastructure operations are in scope, B5 Ops Agent Build™ or C1 OpsCommand™ fit better.
Start with A5 AI Use-Case Due Diligence™ to confirm service desk deflection is the highest-value first move — including a willingness to recommend against it.
Build against the destination platform, not the one being retired. We will say so and recommend sequencing rather than take the engagement.
Fixed fee.
Milestone-based.
Total engagement fee agreed in the scope statement. Not time-and-materials. Not day rate. Every engagement is preceded by a scope conversation to ensure fit before commitment.
Six, most asked.
Q_01What does Service Desk AI Assist actually deliver?
Q_02Is this a chatbot?
Q_03How do you prevent bad answers reaching users?
Q_04What is KnowledgeOps and why does it matter?
KnowledgeOps is the operational model that keeps service desk knowledge current with named ownership per knowledge domain, SLAs for review cadence, and audit trail for knowledge changes.
Without KnowledgeOps, AI triage progressively degrades as knowledge stales — the AI triage that worked at week 1 produces worse outputs at week 20 because the underlying knowledge no longer matches operational reality. B7 delivers KnowledgeOps discipline alongside AI triage so the deflection outcomes sustain rather than decay.
Q_05How do you measure success?
Q_06What comes after B7?
One name
on the engagement letter.
A named Practice Lead is accountable for delivery, commercial outcomes, and the client relationship throughout the engagement. Not a project manager who disappears after kickoff. Not a partner who nods at the SOW and vanishes.
Practice Lead — AI
Present at every phase gate, every scope decision, every difficult conversation. Available for 30/60/90-day post-handover check-ins as part of the engagement.
Including scope amendments.
Signs off all 4 deliverables.
With executive sponsor.
Authorised to negotiate.
CEO within 24 hours.
30/60/90-day check-ins.
Prior. Peer. Next.
AI Use-Case Due Diligence™
3-week due diligence that validates whether service desk deflection is the highest-value first move — and includes the willingness to recommend against it.
Ops Scorecard™
Prior 2-week assessment that identifies AIOps and service desk automation candidates, feeding the B7 execution backlog.
OpsCommand™
Managed IT operations retainer. Operates the service desk stream alongside MTTR, patch/change governance, and availability against SLA.
Thirty minutes.
No slide deck.
A structured 30-minute scope conversation with the Practice Lead. You describe current ticket volume, deflection history, and the state of your knowledge base. We describe whether B7 is the right engagement — and if not, what is.