Between operations and the contract
DIMENSIONAL is a commercial and operational specialist for companies and BPOs putting AI into contact centres.
Our work sits where operations and the contract meet: defining what a resolution is, proving it from your own systems, and designing the human side of an automated operation so the contract can be defended and delivered.
What we are not
Not a platform vendor. We don't sell the bot, and we have no platform to push.
Not a generic AI agency. We don't build chatbots or write general AI strategies. We work on contact centres and the contracts behind them.
Our boundaries
We define what "resolved" means operationally, what evidences it, and how it could be gamed.
Your counsel decides whether a clause is enforceable. Our drafting is commercial input, not legal advice.
Our numbers mean what they say. An unblocking ratio is not a payback. A residual headcount is not a saving.
The work
The assessment runs as a sequence. Each module takes the output of the one before it. The goal is a contract you can defend and an operation that can deliver it.
What could be automated. The first module produces an exposure map and scores how verifiable each segment is. A segment that can be automated but cannot be verified cannot be priced per resolution. Under seat pricing, automation destroys margin.
Resolution specification. Deflection is not resolution. Each automatable segment gets a definition of Resolved that can survive a client-side audit. Every definition needs a positive event in a system of record, and a repeat-contact window that reverses it if the customer comes back. A contained contact where the customer gave up is not billable. Some segments, such as pre-sales and product information, have no verifiable resolution event and are kept out of per-resolution pricing. The output includes commercial drafting for your counsel. It is not legal advice.
Measurement and evidence. The gaps become a build backlog ordered by billings unblocked per day of effort, not by what IT already wanted to build. The usual blocker is not missing data. It is the missing link between the conversation and the system of record. Items that stop over-billing, such as negative containment and honoured requests for a human, go in the first wave even though they unblock no revenue. The ordering ratio is an unblocking ratio, not a payback. Done properly, this also produces most of an EU AI Act deployer evidence pack. It is not a compliance product.
Escalation and handoff. This designs which contacts cross to a human, on what trigger, with what context, and how many people the residual queue needs. Handle time goes up after automation, because humans are left with the hard contacts. A queue that is too small to hold an answer time gets consolidated on shared agent skill, not shared intent. Designed escalations, failures, and customer requests for a human are reported separately. A request for a human is not an AI failure, and it must be honoured. The headcount figure is what the residual queue needs. It is not a saving.
What follows. The residual headcount feeds a workforce plan. The billings unblocked by the evidence work become revenue only if the contract is repriced. The compliance evidence adds little extra effort, because most of it already exists.
How we work
Operations owns the queue and the staffing. Commercial and finance own the contract. IT or data owns the systems of record. Your counsel decides what is enforceable. We do not sell, build, or resell the AI platform.
Every operation is different. We scope the assessment, and its pricing, in the first conversation.