The questions buyers actually ask,
answered straight.

We run enterprise operations — customer service, order and claims, vendor management, finance operations — with AI agents and your own people on the same queue, to the same standard. Your systems of record stay authoritative. We run the work between them, and the business outcome goes in the contract.

A platform, with the outcome in the contract. Forward-deployed engineers connect your systems and get the first workflow live. After that, your own operations lead runs it day to day — no engineering team needed.

No. A copilot suggests and a person still does the work. Here the work is the product: agents run real queues, a supervisor holds the exceptions, and the resolution rate is what we are measured on.

CEOs and COOs of operations-intensive businesses from $50M revenue upward — in retail and e-commerce, industrial and manufacturing, and regulated service operations. The work repeats, crosses departments, and has always scaled with headcount.

No — that is who we are built for. You need a mandate to make AI real and a number you want moved, not an AI team. The platform arrives already knowing how your industry runs, and your own operations lead runs it.

No. CygnusAlpha sits on top of your systems of record, never instead of them. They stay the source of truth; we read from them and write back.

Yes. Approval points are part of the rulebook, not an afterthought. People and agents share one queue on a supervision desk, and any action you decide needs a person gets one.

They are your rulebook — your entities, thresholds, policies, approval limits and escalation paths, written as structure rather than as prose inside a prompt. The operations lead who owns the number edits them directly. Versioned, approved, no code and no engineering ticket.

The state cannot change except by a permitted move. An unauthorised action is not discouraged — it is refused, logged, and the agent is handed the moves it may make, so it carries on rather than stalling.

It escalates to a named person, the correction is captured from that person’s decision, and the rulebook learns from it. Nothing learns from an unchecked call.

Yes. Every action is recorded with who took it and on what evidence. The record is append-only — it only ever grows, and nothing in it is rewritten.

Yes. Every agent and human action, the evidence behind it, and the rule that permitted it. That record is the point, not a feature bolted on afterwards.

Three ways, like a new hire. Live context that expires on a date. Corrections captured from a supervisor’s decision. Reflections distilled overnight. Every entry is activated by a named person.

Yes. Every entry is stored as readable text with its provenance — a sentence someone can check, not weights nobody can inspect. Superseded entries retire rather than being overwritten.

Yes, and the asset compounds. Agents are a commodity; your rulebook is only yours and gets deeper every month. That is what makes an account hard to unpick after a year.

Yes. Single-tenant VPC in your own account on Google Cloud, Azure or AWS.

Yes. Inside your perimeter, with no external connectivity, for operations where isolation is the requirement.

Yes. Multi-tenant SaaS in our cloud, with no capital expense, is the fastest start.

Wherever you choose it to — our cloud, your cloud account, or inside your perimeter. Your systems of record keep the records either way; we do not become the system of record.

You do, from day one, contractually. Workflows, integrations, business policies, escalation hierarchies, your data and your memory — in your tenant, exportable in open formats.

Everything configured for you is yours from day one. Deployed on your infrastructure the software keeps running, with source held in escrow and defined release triggers. And after implementation the platform can be yours to keep and run under a perpetual licence.

ISO 27001:2022 and ISO 9001. Two provisional patents filed. CygnusAlpha, Inc. is a Delaware C-Corporation.

ISO 27001:2022 today. SOC 2 Type II is in progress. If your process needs the report in hand before signature, tell us at the first conversation and we will sequence around it rather than waste your time.

Every engagement starts from a measured baseline and an agreed target KPI. We operate the workflow against that target. If the metric does not move, we go back in at our cost.

It is scoped to the metric agreed in advance — not a warranty on your whole P&L. The exclusions, and the customer-side dependencies that suspend it, are written into the contract before either side signs.

Three components. Build — one-time, fixed price, and we carry the scope risk. Licence and operations — annual, the platform plus the capacity running the operation against the agreed metric, banded and never priced per head. Usage — metered per business interaction, where an interaction is a complete business event rather than a compute event.

No. Pilots are paid, time-boxed to one measured outcome, and credited against year one. A free pilot buys attention rather than a decision, and it is the wrong start for work we are going to be accountable for.

Weeks, not quarters. In a sector where the rulebook is already templated, the first workflow goes live four to six weeks from contract.

References are available on request. Our production deployment has been live eighteen months across eleven workflows, with resolution without a person moving from 20% to 68% and cost to serve down 97%.

Thirty minutes about fit: how the operation runs today, where the cost sits, and whether the number you want moved is one we can be accountable for. No demo, no audit, no diagnostic.

A number you want moved, the person who owns it, and access to the systems the work already touches. Not an AI team, and not a data project first.

The systems the work already crosses — CRM, ERP, ticketing and service desks, billing, warehouse and inventory, and the messaging channels your customers use. They stay authoritative; we read from them and write back.