Six ways to put AI into an operation.
Ask where each one stops.
Most comparisons line up vendors. This one lines up models — because the model decides who is left holding the outcome when the project ends.
In one line
Every model below does something well. Only one is contractually tied to the business number the operation exists to move — and stays until it moves.
Not a platform you staff. Not a project with an end date. Not headcount.
The six models, side by side
Scored on the six questions that decide who owns the result. Categories, not companies — the best vendors in each do their job well.
| Model | Who builds it | Who gets it live | Who runs it day to day | What it is measured on | Outcome in the contract | Where the obligation ends |
|---|---|---|---|---|---|---|
| AI agent platform | You do, with their tools | Setup services, often | You do | Resolutions, in the channel it covers | Sometimes priced per resolution | At go-live |
| In-function agents — inside your CRM or service desk | Configured inside one system | Inside that system | Inside that system’s walls | That system’s own metrics | No | At the edge of its own system |
| Systems integrator | Yes | Yes | Rarely — or as staff, per head | Milestones delivered | Rarely | At invoice |
| AI build shop | Yes | Yes | No | The build shipped | No | At handover |
| Outsourcer | People, more than systems | Yes | Yes — their people | Service levels | Service levels, priced by headcount | When the contract does — and the cost scales with the queue |
| CygnusAlpha | A platform that arrives with your industry in it | Forward-deployed engineers | We do, your team does, or both | Your business KPI, from a measured baseline | Yes — if the number doesn’t move, we go back in at our cost | When it’s real — and it can be yours to keep |
The question that sorts them
Where does the obligation end?
The platform vendor’s ends at go-live.
Powerful tools, handed over. Integrating them across your systems, and running them on Monday morning, is yours.
The integrator’s ends at invoice.
The project is delivered to specification. Whether the resolution rate moved in month four is nobody’s line item.
The outsourcer’s never ends — and neither does the headcount.
The work gets done. The cost line rises with the queue, the way it always has.
Software is accountable for software. Delivery is accountable for delivery. Nobody is accountable for the outcome — that is the gap we are paid to close.
When another model is the better choice
We would rather tell you now than find out in month four.
| If… | Choose |
|---|---|
| You have a strong engineering team and one channel to automate | An agent platform. You will build it and run it well yourselves. |
| The work never leaves one system | That system’s own agents. Work that crosses departments is the problem we solve; if yours doesn’t cross, you don’t need us. |
| You need a one-off build, and people to run it afterwards | A build shop. Cheaper for a project with an end date. |
| You need an enterprise-wide, multi-country transformation programme | A large systems integrator. That is a different job, at a different scale. |
| You want the whole function run by someone else’s people | An outsourcer. With us the org chart doesn’t change — the team does. |
| Your security review needs a SOC 2 Type II report in hand before signature | A vendor that holds one today. Ours is in progress; ISO 27001:2022 is certified. Tell us at the first conversation and we will say honestly whether we can sequence around it. |
Where we fit
An operations-heavy business with a number to move, a mandate to make AI real, and no AI team to staff it. The work crosses departments. It repeats in volume, not milestones. And its cost has always risen with headcount.
A number to move. A person who owns it. No AI team required.
Every capability, with an explicit yes or no — including the ones we do not offer. Platform capabilities →
Our cloud, your cloud account, or air-gapped — with your data in your tenant either way. Deployment and data residency →
Thirty-two answers on approval, rules, learning, data and commercials. Read the FAQ →
If you are choosing between models, start here.
Thirty minutes about fit — not a demo, not an audit, not a diagnostic. If another model suits you better, we will say so.
CygnusAlpha. We turn AI ambition into AI operations.