Run AI. Not an AI department.
RAILS is the client portal that comes with every OPINOV8 AI Agent delivery. One place to see what’s running, know what’s working and prove what AI delivers — and it lives inside Microsoft Teams, so your people don’t have to battle with yet another tool.
The hyperscalers will tell you tokens burned and uptime. RAILS and OPINOV8 tell and measure the things they do not: hours returned, value delivered and every decision on the record.
Proof your CFO will actually read.
Translate AI activity into hours returned, cost saved and measurable operational value.
Most AI never leaves the sandbox. RAILS counts the days until yours does.
RAILS tracks every calendar day from idea capture to the first production run. The 12-day median separates four days of active work from eight days waiting for decisions, access or approvals.
Median idea to live
Working / waiting
- Start2 days
Capture & proposal
The business owner describes the job and agrees the outcome.
- G14 days
DPO check
Data use, compliance and DPO approval.
- G21 day
Connections
Connections approved by IT.
- G33 days
Prototype
Runs safely with real data in the sandbox.
- G42 days
Production
First live run, with value measured from day one.
Live portal — Time to live, 7 agents Mar–Sep 2026 · click to enlarge
See it. Run it. Prove it.
Real screens from the live portal — no mock-ups.
See it
Describe the job in plain words — or drag and drop a document with an operational procedure you want to automate. RAILS maps the sequence against your connected systems while you watch.
Idea wizard — from a sentence to a full sequence
Six views. One source of truth.
From the first idea to the controls that keep it safe — explore the real portal.
Designed by the people who own the process.
A business stakeholder describes the job in plain language. RAILS maps the agent's interactions across departments, systems and client touchpoints — and every message is reviewed and editable before anything sends.
- Plain-language brief, no technical spec
- Interactions mapped across teams and clients
- Every customer message reviewed before send
Live portal — Ideation Wizard · click to enlarge
AI handles the routine. People handle the exceptions.
Every agent carries a named lead, a confidence threshold set at G1, and an approval route declared before it ever runs. Anything below the threshold — or outside policy — stops and goes to a person.
And escalation is not failure — it is the control working. Each one carries a reason a human can read, and the whole estate runs at 2.5% against a 5% design point.
Every escalation is dated, named and on the record — so the number can go down next month for the right reason.
One live agent, step by step · click to enlarge
180 approval decisions on a single agent in September — a designed step in the sequence, not an accident.
Out of policy, below threshold, data error — the reason is written down next to the decision, every time.
When the agent cannot check, it stops. No answer, no send — it never guesses its way through.
What is running, when, and why.
Most businesses cannot answer that about their AI. RAILS splits the estate into three plain groups — agents your people run themselves, agents on a cadence, and agents that only fire when something happens outside RAILS.
On-demand is the part people underestimate. It is a toolbox: the sales report is scheduled for Monday, but you need it on Wednesday for a board call — you run it yourself, in seconds, with the same guardrails and the same record as everything else.
- Every live agent in one inventory, grouped by how it starts
- Run any on-demand agent yourself, whenever you need it
- Cadence and next run visible at a glance on the calendar
- Value shown per agent, so you know what each one is worth
On-demand — your toolbox
Agents your people run themselves, whenever they need them. The board pack appendix, the aged debtors snapshot, the competitor digest — each one with its last run, who ran it and what it is worth.
Change is a constant, trust is not.
“Production-grade” is not a bar an organisation clears once. It is a ladder each workflow climbs on its own — level by level, with evidence at every step. The level an agent sits on today tells you exactly what you are trusting.
Data drifts, processes change, models update. These indicators exist so trust is adjusted continuously — not reviewed once a year at an audit.
- Level 1
Information
Deterministic retrieval and calculation across your systems. Correct, but mute — a faster query, not a colleague.
You trusta fetch - Level 2
Insight
The agent explains what the numbers mean. This is where AI enters — you are trusting an explanation, not just a fetch.
You trustan explanation - Level 3
Prediction
The agent forecasts what is likely next. Trusting a forecast is a bigger ask than trusting an explanation of the past.
You trusta forecast - Level 4Human gate
Prescriptive
The agent drafts the action — an email, a journal entry, a record update — and a named person decides whether it goes.
You trusta drafted action - Level 5Audited
Autonomous
The agent decides and acts within a pre-trusted scope, above a confidence threshold — audited, and reversible by a named human.
You trustthe action itself
One company. Many levels at once.
There is no company-wide maturity score. A finance reconciliation agent can sit on level five while a sales outreach agent in the same portal sits on level four — simultaneously. RAILS sets confidence thresholds and approval routes per agent, not per client.
Levels move. In both directions.
An agent that earns autonomy can lose it when the process changes — and that is the point. RAILS shows which level every agent is on today, so scope is widened on evidence and tightened the moment something shifts.
- Up on proven runs, not promises
- Down the moment data or process drifts
- Every move dated, named and on the record
OPINOV8 anchors and guides. RAILS keeps the record.
Deciding which level a workflow should reach next — and when it is safe to get there — is judgement work. That is what we do with you, agent by agent.
OPINOV8 — the guide
Our delivery team sits on your side of the ladder: choosing the next level, configuring the thresholds and approval routes, and widening scope only when the evidence supports it.
RAILS — the anchor
Every level, every move up or down, and every decision in between lives on the record in your portal — so trust is something you can see, not something you are asked to assume.
Most people hear “guardrails” and think technical.
We have the technical side covered, obviously. But there is much more to it: clear ownership of every agent, escalation rules when something looks wrong, risk thresholds that decide what goes out on its own and what waits for a named person.
So when the auditor turns up and the EU AI Act is in force, there is no panic — and no need to call OPINOV8. Full compliance documentation and a complete audit trail are built into everything you run on RAILS. Not an afterthought.
- A named owner and reviewer on every agent
- Escalation rules and confidence thresholds set per agent
- Every gate approval immutable, dated and attributable
- Compliance documents maintained automatically as you run
Ownership and oversight
A named lead, a confidence threshold and every review decision counted — approved, amended or rejected — so you always know who is accountable and what they decided.
No new tool to learn. It lives in Teams.
All user and agent interaction can be embedded directly into Microsoft Teams. The general workforce meets AI where they already work — approvals, nudges and results arrive in the flow of their day, with nothing new to sign up for or onboard onto.
Included with OPINOV8’s AI Agent delivery — RAILS is how our clients see, run and prove every engagement. It is never sold as a standalone portal or an extra licence.
Talk to agents in chat
Requests, updates and answers happen in conversation, inside the Teams channels people already use.
Approvals come to you
“Needs your attention” cards land with the right person in the right channel — one click to review.
Zero onboarding
The wider workforce never adopts another tool. If they have Teams, they already have RAILS.
Your agents, built with us, wherever they run.
RAILS is how Opinov8 co-creates agents with your teams. It’s the shared portal and operating model that takes an idea to production and keeps it accountable once it’s live. It works with any agent harness, so your agents don’t have to move to work with us.
Built and run by Opinov8
We build your agents and run them on our own harnesses, with governance, monitoring and business metrics in place from day one.
Built in your Microsoft stack
Already working in Copilot Studio, Microsoft Foundry or your own framework? We connect RAILS to your existing tenant. Your agents keep running where they are, and we bring them into the same governance and the same metrics.
Or a mix of both
In-house agents and agents we deliver sit in one estate, held to one standard.
Wherever an agent runs, RAILS gives both teams the same answers:
What should we build next?
Ideas come from the business, and we scope and prioritise them together.
Is it safe to go live?
Every agent passes the same gates, with named approvers on your side and ours.
Is it working?
Value, adoption and risk are measured across the whole estate, not agent by agent.
Where is it stuck?
You see where time to production goes and who it’s waiting on.
One way of working for every agent. Ideas move from first request to production faster, and both teams always know what’s live, who owns it and what it returns.
You don’t need RAILS or Opinov8.
If you want a personal productivity agent — one person, one task, one tidy summary — you genuinely don’t. Build it in an afternoon with whatever tool is closest to hand.
You don’t need RAILS or Opinov8 when…
- …it’s a personal productivity agent for one person.
- …the output is a note to yourself.
- …nobody else has to see it, approve it or trust it.
You definitely want it on RAILS when…
- …it’s a core business process, not a side experiment.
- …it crosses departments and touches clients at multiple points.
- …your IT department is always busy — a bottleneck.
- …it takes forever to execute stuff through your internal ticketing process.
- …the output has to be on the record.
That’s when you put it on RAILS.
Watch one rail, end to end.
From an idea typed in chat to a live, sandboxed workflow with a human gate — the actual portal, not a render.
Every AI Agent we build comes with RAILS.
You don’t buy RAILS — it comes with us delivering your AI. Bring one process to a short, guided session and see your own workflow running, measured and proved.