Reads the platform
Queues, skills, agents, opening hours, wrap-up codes, prompts, entry points and flows, over the vendor's own API. Reads only. It writes nothing to the source and is safe to run against a system in production.
Capgemini
Read a client's Genesys, NICE, Talkdesk or ServiceNow configuration, see exactly what can be migrated to Dynamics 365 Contact Center and what has to be rebuilt, and write nothing until somebody has approved the plan.
What it does
A migration is not one job. It is reading a platform honestly, deciding what can go, and writing it somewhere a person has agreed to.
Queues, skills, agents, opening hours, wrap-up codes, prompts, entry points and flows, over the vendor's own API. Reads only. It writes nothing to the source and is safe to run against a system in production.
Every entity is graded by what the source gives up and what Dynamics accepts, and the worse of the two is what actually happens. Anything that cannot move is named, one record at a time, with an estimate against it.
A plan lists every record it would create or change. Nothing is written until a named person approves that exact plan, and every write is recorded before and after so it can be undone.
Platforms
Seven contact centre platforms, described by how much of each one this can actually read. The number is counted from the capability contract the product itself uses, so this page cannot claim more than the tool delivers.
reads 23 of 37 entity types
reads 10 of 37 entity types
reads 14 of 37 entity types
reads 18 of 37 entity types
reads 19 of 37 entity types
reads 20 of 37 entity types
reads 22 of 37 entity types
Not on the list? Any platform can be brought in as a file drop: CSV, JSON or Excel exports mapped through the same wizard. Dynamics 365 is where this writes, and can also be read as a source when a migration is a consolidation.
What it knows
Every capability claim carries a verification state. Documented means the vendor published it. Inferred means we worked it out. Verified means somebody reconciled it against a real environment on a recorded date, and the product says which of the three you are looking at.
That distinction is not academic. A wrong table name fails loudly and a wrong column name fails loudly. A right column name that means something else does not fail at all, and only reading a real environment finds it.
Safety
Discovery, normalisation, analysis, reading the target, mapping and planning are all read only. Exactly two stages write, and both sit behind an approval recorded against a named person and the exact plan they saw.
There is no setting that skips it. A tool that can write four hundred routing records into production unattended is a tool that will, once.
The approval records a hash of the plan that was read. If the plan changes afterwards it no longer matches and the apply refuses, because approving one thing and applying another is the failure the gate exists to prevent.
A record that already exists in the target and that this tool did not create is a conflict. Nothing is overwritten on a name match alone, and deciding whether it is the same record is a person's job.
What is about to happen and what is being replaced are recorded before the request goes out, and the outcome afterwards. Without the before state there is no rollback, only hope.
Ten consecutive failures, or a fifth of at least twenty attempts, and the run stops itself. A wall of failures is the environment telling you something, and pushing through only makes the cleanup bigger.
What you get
A screen is where findings are read. These are what gets forwarded, filtered, argued with in a workshop and pasted into a statement of work, and they have to stand up with the product nowhere near them.
Excel. Every entity with its record count, what the source gives up, what Dynamics accepts and the effort range for each. Plus the gap analysis, the risks and the target mapping. Every sheet carries its own caveat, because a sheet gets copied out of a workbook on its own.
What to do before the day, on the day, and when it goes wrong. The write order comes from the model's own dependency graph, so the runbook cannot list the steps in a different order from the one the tool uses.
Everything no tool can migrate, one record at a time, with the evidence and an estimate. "Flows are manual" is not a finding. Forty-one named flows with hours against them is what a client is buying when they ask what a migration costs.
The product
Six languages, light and dark, engagements with roles, and a demonstration estate everybody can open on their first sign in.
Languages
The interface, the guidance and the reports, in all six. The page you are reading is one of them, and the chooser in the corner keeps your place rather than sending you back to the top.
Limits
A page that only claims is a brochure. These are the things this tool is honest about not doing, and they are the things that take the time.
No tool does, from any source, to this target. Flows are read and parsed into a neutral vocabulary so they can be described and counted, and then they go to the backlog with a diagram and an estimate.
Users in Dynamics come from Entra and a licence. An agent who cannot be matched to a real person is reported, not created, and their skills wait for somebody to say who they are.
Every estimate is a range, and the range widens when the target names have not been checked against a real environment. There is nowhere in the product to put a single figure, which is deliberate.
Run a discovery against the client's platform. It reads and writes nothing, it needs no Dynamics environment, and at the end of it you have a workbook, a runbook and a number with two ends.