Enterprise implementation
From the process you run today to an AI workflow you can accept.
Most organisations need one process that works before they need an AI strategy. We start from how the work is actually done, agree what the first implementation must do and how it will be accepted, and then build it on the systems you already use.
From process to implementation
1. Understand the process
We meet the people who do the work and go through real cases: what arrives, which files and systems are used, where information is missing, where judgement is needed and who approves. Mapping across several departments usually takes one to two weeks, depending on people’s availability.
The result is written down: a map of processes and information sources, a summary of needs per department, a prioritised list of opportunities, and a list of gaps that need a decision.
2. Prioritise
Candidates are compared on business value, complexity, the systems involved, data availability and dependencies. Some processes are better served by adopting a tool you already own under a clear procedure than by building something new. Others are left out, and the reason is recorded.
3. Information sources
For each process we list the files, spreadsheets, boards and systems it depends on, their formats, how often they change and where they disagree. Rules for missing or unusual values are agreed with the process owner. A zero, for example, is not automatically treated as an error.
4. Integrations
We connect to the systems already in use, such as ERP, CRM, project boards and document stores, through their interfaces where possible. Existing systems remain the system of record. Any use of a third-party AI service with your data is agreed in advance.
5. Human checkpoints
We decide with you where a person must review before work moves on: an unusual result, a negative commercial outcome, a contract deviation. These cases are flagged with the supporting data and are not resolved automatically.
6. Testing
Before operational use, results are tested with the process owner on agreed samples and compared with the current method. Open gaps, permissions, who approves and who operates the workflow are confirmed. Operational use begins only after explicit approval.
7. Adoption
Training is given by role, on the processes each person will use, with short working guides. After launch there is a defined support period for questions and fixes.
Case study
Optimising workflows across several departments
Client project
Liba worked with a large enterprise to optimise workflows across many departments. Each department had its own way of working, so instead of forcing one process on everyone, each gets a tool shaped around its own methodology, and all of them talk to each other and share the same data.
Turning raw customer data into a ready-to-use profile
Before a new offer, the commercial team needs a clean profile of the customer’s data.
Who decidesAn analyst reviews the profile before it is used.
Tracking contract commitments across projects
Who decidesThe project manager checks every deviation and decides what to do.
What affects feasibility, scope, cost and timing
- How clearly the rules can be written down, and who has the authority to approve them.
- The condition of the data: formats, gaps and how many sources must be reconciled.
- How many systems are involved, and whether they have usable interfaces.
- Security and permission approvals on your side, and how long they take.
- How many human checkpoints and exception routes the process needs.
- Whether the goal is a demonstration, a pilot or operational use.
- Whether the people who know the process are available for mapping and testing.
Before each implementation we write a scope document covering the objective, inputs and outputs, systems and data, effort, dates, success measures and acceptance conditions. Scope, timing, deployment and ongoing support depend on what is agreed in it.
Questions and answers
How do we choose the first process for AI automation?
Choose a process that is frequent, rule-heavy and time-consuming, with a clear owner and data you can reach. Avoid one whose rules nobody can state or whose main systems cannot be accessed. Our guide sets out the criteria in more detail.
What does an implementation project include?
Mapping, a written scope with acceptance criteria, building the workflow and its connections, defining human approvals and exceptions, testing with your process owner, training and handover. Each is a deliverable you can review.
What affects the timeline and cost?
Mainly the clarity of the rules, the state of the data, the number of systems and how quickly access is approved. We give an estimate after mapping and fix it in the scope document. First processes are planned at roughly five to nine weeks including adoption. Yours may differ.
How do we verify that the workflow works correctly?
Baselines are agreed with the process owner before the build: for example preparation time, how many missing values or exceptions are caught in a sample, and agreement with the approved calculation method. The workflow is tested against those baselines and accepted when it meets them.
Discuss a process you want to improve.
Describe the process, its main difficulty and the systems it touches.
Discuss a process you want to improve