What an Enterprise AI Workflow Audit Costs
Concentric Lab's two-week AI Workflow Audit costs $10,000, credited toward the build. See its scope, outputs, exclusions, and decision criteria.
Concentric Lab's AI Workflow Audit costs $10,000 and runs for two weeks. If the client proceeds with the build, the full fee is credited toward it. That is the complete published price of the audit we sell today. The boundary is the workflow, not an invented enterprise tier. One recurring operation must be narrow enough to inspect, test, and turn into a build decision inside two weeks. The price does not buy a list of possible use cases. It buys a decision about the first system worth funding. A useful audit may recommend a build. It may also recommend assist only, keep the decision human, or defer the work until the data and access exist. If every path ends in a proposed project, the audit is sales qualification with a document attached. The Operational Design Session is the public starting point. A prospect describes one recurring process. The session maps the work, asks for the facts that change the design, separates system actions from human decisions, and recommends a first-release boundary. The completed brief includes a workflow map, first-release scope, human checkpoints, proof criteria, an operating boundary, dependencies, blockers, failure paths, and acceptance tests. That session does not replace the audit. It tests whether the problem is concrete enough to investigate and makes the first conversation specific. It cannot verify system access, inspect the actual records, interview the operators, or settle conflicting rules inside the company. The paid audit begins where the public session stops: evidence inside the operation. Current workflow map. We document the trigger, every material step, the systems and people involved, the output, the known exceptions, and the point where a case stops moving. This is the workflow as it runs, not the policy as it was written. Economic baseline. We capture the hours, delay, rework, error exposure, and constrained volume attached to the current process. Each candidate system is priced against the cost of doing nothing. Where a number is not available, the plan names the missing input instead of manufacturing precision. First-release scope. The plan states what to build now, what can only assist a person, what should stay human, and what should be deferred. It defines the smallest release that produces an operational result rather than a demonstration. Operating boundary. The brief states what starts the system, what inputs it requires, what it produces, and what happens when it cannot proceed. System responsibilities and human decisions appear separately. This is where a workflow becomes safe enough to estimate. Controls and proof. The plan names human checkpoints, failure paths, guardrails, dependencies, blockers, and acceptance tests. Proof criteria describe what has to be true before the release can be trusted with real work. A production workflow needs a way to stop, a person who can decide, and evidence that the decision is improving. Portable build plan. The client can take the result to an internal team or another vendor. The fee is not a deposit that only becomes useful if Concentric Lab receives the build. If we build it, the $10,000 is credited in full. The audit is not the production build. It does not connect every system, migrate historical data, operate a model against live decisions, or replace the people currently running the process. It does not survey every department for possible automation. Broad inventories produce long lists and weak priorities. We start with the recurring operation a buyer can name and the constraint that makes it expensive. It also does not guarantee that a build is justified. The result can be defer when the inputs are inaccessible, keep human when the judgment cannot be bounded, or assist only when the system can assemble evidence but should not decide. Those are useful outcomes. They prevent a $150,000 to $400,000 build from being aimed at a workflow that cannot support it. The build is priced after the audit. Concentric Lab's fixed-fee builds typically run $150,000 to $400,000 over 8 to 16 weeks. The range moves with scope and integration depth, not with how impressive the model sounds. A workflow that reads one stable system, proposes an action, and waits for approval is smaller than one that writes across finance, customer, and operations systems while preserving permissions and an audit trail. Five conditions expand the work. More systems require more integration and more failure handling. Unstructured or inconsistent data requires extraction and reconciliation before decisions can be trusted. Higher-consequence actions require stricter approvals, testing, and auditability. A large exception surface requires better routing and more operator tooling. A weak handover path requires documentation, training, monitoring, and an evaluation harness that the client's team can run without the builders. The model is not the cost center. On our current builds, the model itself accounts for about 20 percent of the cost. Architecture, integration, and the evaluations that keep the system reliable in production account for the rest. A cheap prototype can prove that a model can produce the answer. It does not prove that the surrounding operation can trust, route, reverse, and measure that answer. Run the public Operational Design Session when the problem is still loosely stated. Use it to see whether one recurring process can be mapped and whether the first boundary produces a credible recommendation. The cost is zero, and the output is directional rather than verified inside the operation. Buy the two-week audit when the workflow is real, recurring, and expensive enough to investigate, but the organization has not settled what to build. The audit is the right unit when the buyer needs a verified scope, economic baseline, human boundary, controls, and acceptance tests before approving a larger commitment. The price is $10,000, credited if the build proceeds. Move directly to an internal build when the workflow, access, owner, controls, and proof criteria are already explicit. In that case the organization does not need an audit from us. It needs delivery capacity. If the internal team has that capacity, keep the work internal. If it does not, the build can be scoped from the existing specification. Do not buy the audit when the objective is to find somewhere to use AI, when no operator owns the process, or when the required systems cannot be accessed. Those are not scoping problems. They are sponsorship and access problems. Two weeks can identify them. It cannot solve them on behalf of the person who holds the authority. At the end of the audit, the buyer should be able to answer five questions without interpretation. What is the first release. Which decisions stay with people. What can stop the system. What evidence proves it works. What does the status quo cost compared with the proposed build. If one answer is missing, the project is not ready to price. Our answer is delivered as a prioritized build plan with the ROI math. It belongs to the client. If the recommendation supports a build, Concentric Lab can architect and ship it inside the client's repository and cloud. Typical builds run 8 to 16 weeks at a fixed fee. After release, we can operate the system from $3,000 a month, or the client's team can take it in-house with the runbook and operators trained during the build.