Retainers

Managed AI Operations

Your deployed AI systems, kept healthy: monitored, tuned, and governance-reviewed every month.

The engagement

A monthly retainer for organisations running AI systems in production — roughly 8 senior hours per month spent monitoring uptime, cost, and quality drift, tuning prompts and configs against real usage, and keeping a governance-aware paper trail current. You get a live operations dashboard, a monthly written report, a quarterly governance review, and a reactive WhatsApp/Slack line when something breaks. Built for the period after launch, when an AI workflow quietly stops behaving and nobody notices until it costs you.

What we review

  • Monitoring deployed AI systems: uptime, cost, and quality drift

  • Monthly prompt and configuration tuning

  • Quarterly governance review and risk-documentation upkeep

  • Operations dashboard and monthly reporting

  • Reactive WhatsApp/Slack incident support

Why deployed AI needs ongoing support

Most AI systems do not fail loudly. A prompt that worked at launch slowly drifts as your data changes; an API price tier shifts and your monthly cost doubles without warning; a model update quietly changes how a workflow answers, and the first you hear of it is a complaint from a broker, a patient coordinator, or an analyst. Without someone watching cost, quality, and the governance paper trail, a working system degrades into a liability — and in a UAE/GCC context, that liability is both operational and data-protection-aware.

  • Quality drift: outputs that were accurate at launch slowly degrade as inputs and models change

  • Cost creep: token and API spend climbing unnoticed until the invoice arrives

  • Stale governance: risk documentation, vendor lists, and approval workflows falling out of date as systems evolve

  • No owner: the person who built it has moved on, and nobody is accountable for keeping it healthy

What is included

A fixed monthly block of roughly 8 senior implementer hours, spent on your live systems rather than meetings. The work splits across monitoring, tuning, governance upkeep, and reactive support — weighted each month toward whatever your systems actually need.

  • Active monitoring: uptime, cost, and output quality tracked continuously, not spot-checked

  • Monthly tuning: prompts, retrieval configs, and thresholds adjusted against real usage data

  • Operations dashboard kept live with cost, volume, error rate, and human-approval rate

  • Monthly written report on what changed, what drifted, and what we recommend next

  • Quarterly governance review of risk documentation, vendors, and human-approval workflows

  • Reactive WhatsApp/Slack line for incidents and quick questions

What is not included

This is a maintenance and governance retainer, not a build retainer or a law firm. We keep what you have running well and your risk documentation current — we do not ship net-new systems under this tier, and we do not certify, guarantee, or provide legal advice on any regulatory matter.

  • Net-new builds — scoped and quoted separately under the Build tier

  • Tool, model, and API subscriptions — billed to you directly by the vendor

  • Compliance certification or legal advice — we are governance-aware, not a law firm; we keep documentation current and do not guarantee any regulatory outcome

  • Hard SLA / 24-7 on-call — reactive support is best-effort within business hours

The monthly rhythm

The retainer runs on a predictable monthly cadence so cost and attention are both predictable. You always know what was watched, what was changed, and what is coming.

  • Continuous: automated monitoring on uptime, cost, and quality drift, with alerts on anomalies

  • Mid-month: prompt and config tuning applied against the month's real usage data

  • Month-end: written report and an updated dashboard delivered to your operations lead

  • As needed: reactive WhatsApp/Slack support for incidents between scheduled work

  • Quarterly: a deeper governance review layered on top of the normal month

Quarterly governance review

Once a quarter we step back from day-to-day tuning and review the system against its governance posture. This keeps your risk documentation, vendor list, and human-approval workflows aligned with how the system actually behaves now — not how it behaved at launch.

  • Risk documentation refreshed to match the system's current behaviour and data flows

  • Vendor and model changes reviewed for data-protection-aware and PDPL-aware implications

  • Human-approval workflows checked: are the right decisions still routed to a person?

  • Audit trails verified so AI decisions and approvals remain logged and traceable

  • A short governance memo your leadership can put in front of a board, investor, or regulator-facing stakeholder

Dashboard and reporting

You get a live operations dashboard plus a monthly written report, so AI is no longer a black box your team hopes is working. The dashboard answers the operational questions; the report adds the judgement.

  • Dashboard: cost over time, request volume, error rate, latency, and human-approval rate at a glance

  • Quality signals tracked so drift is caught early, not after complaints

  • Monthly written report: what we changed, what drifted, what we recommend, what to watch

  • Plain-language summaries an operator or founder can read in five minutes — no jargon required

Use cases

Real-estate brokerage

Keeps an inquiry-triage automation and listing-drafting workflow tuned as portals, lead volume, and bilingual EN/AR phrasing shift, with approval steps and audit trails kept intact.

Clinic / healthcare admin

Monitors a patient-comms or scheduling assistant for quality drift and cost, with quarterly review of data-protection-aware handling and human-approval points.

Fund / finance operations

Watches a RAG knowledge assistant or ops automation for accuracy and cost, keeping risk documentation and audit trails current for DIFC/ADGM-facing scrutiny.

Family office

Acts as the accountable owner for deployed AI tooling — monitoring, tuning, and a quarterly governance memo leadership can rely on.

What you actually get

  • ~8 hours/month of senior implementer time, scoped to your live systems

  • Active monitoring on deployed AI: uptime, token/API cost, and output quality drift

  • Monthly prompt + config tuning based on real usage data, not assumptions

  • Live operations dashboard tracking cost, volume, error rate, and human-approval rate

  • Monthly written report: what changed, what drifted, what we tuned, what to watch

  • Quarterly governance review — risk documentation, vendor changes, and human-approval workflows checked against your current setup

  • Reactive WhatsApp/Slack support line for incidents and quick questions

  • Audit-trail review so logging of AI decisions and approvals stays intact as systems change

Who buys this

  • Operators running one or more AI workflows in production (automation, RAG assistant, ops dashboard)

  • Founders who built or commissioned an AI system and have no one accountable for keeping it healthy

  • Real-estate brokerages, clinics, family offices, and fund/finance teams that need a governance-aware paper trail kept current

What's not in scope

  • New build work or net-new systems (scoped separately under the Build tier)

  • Tool, model, and API subscription costs (billed to you directly by the vendor)

  • Legal advice or compliance certification — we keep your risk documentation current, we do not certify your AI or guarantee any regulatory outcome

  • 24/7 on-call or hard SLA response times (reactive support is best-effort within business hours)

Common questions

Newsletter

One letter, every Sunday. Working systems — not hot takes.

Build logs, working systems, and field notes from running a portfolio of AI ventures. Sent weekly, never more.

Weekly. No spam. Unsubscribe anytime.