An automation assessment is not a software demo. It is a structured review of how work actually moves through the business, followed by a decision about what to automate, what to simplify and what to leave alone. The output should be a ranked set of opportunities with owners, evidence and a cost model, not a list of tools.
This explainer covers the components, how the work is done, a worked example and the limits you should expect.
Why assess before automating
Automating a broken process makes errors faster. If an order is keyed into three systems because no one trusts the first one, a bot can key it into three systems more efficiently, but the underlying disagreement remains. The assessment finds the disagreement before anyone buys a licence. It is also a chance to stop work that should be eliminated rather than automated.
The seven components
- Process inventory. For each process: trigger, steps, owners, systems, volume, cycle time, exception rate and customer impact.
- Friction analysis. Waiting, rework, duplicate entry, unnecessary approvals, handoffs and shadow spreadsheets.
- Data and system readiness. APIs, master data, identity, audit logs, retention and the quality of the input.
- Rule and exception mapping. What can be decided by a rule, what needs judgment, and who handles the exception.
- Suitability scoring. Frequency, rule-based share, data availability, error cost, cycle time and volume.
- Business case. Hours released, error and rework cost, cycle-time improvement, compliance value and run cost.
- Roadmap. Quick wins, integration work, change management and a review date.
How the work is done
Interviews alone produce the official process. To see the real one, observe a sample of work: sit with the team, watch a batch of orders or claims, and collect the documents, emails and spreadsheets used at each handoff. Sample the exception queue, not just the happy path. If the systems keep logs, compare timestamps to find waiting time.
The assessment team should include someone from operations, someone from finance or compliance, and someone who can assess the systems. IT alone will miss the customer impact; operations alone will miss the integration constraints.
A focused assessment for one process usually takes two to four weeks, depending on how many people must be interviewed and how easily the systems can be inspected. A company-wide review takes longer. Agree the scope and the decision it must support before booking interviews, or the work expands into an audit of everything.
A worked example
Suppose a distributor receives 400 orders a month by email, phone and WhatsApp. A sales representative re-keys each order into the ERP, checks stock, asks finance to approve credit, then sends a confirmation. The assessment may find six handoffs and three re-entries per order.
At 12 minutes of re-entry per order, 400 orders consume 4,800 minutes, or 80 hours a month. That is an illustrative calculation, not a benchmark. The value comes from showing where the time goes and what a change would remove. If a structured order form and an ERP integration remove two of the three re-entries, the saving is real only if the exception rate stays low and the team adopts the new flow.
The same example usually reveals a second opportunity: credit approval waits for one person. A rules-based approval for orders below a threshold, with an audit trail, may remove more waiting time than the data entry itself.
What it cannot fix
- Unclear policy. If no one agrees on who approves a refund, automation will only encode the confusion.
- Poor master data. Duplicate customers and items produce duplicate automations.
- Unstructured exceptions. If every case is different, a human still needs to decide.
- Supplier behaviour. You cannot automate a partner who sends a different PDF every week without an extraction and review step.
- Adoption. A workflow that saves the manager time but adds work for the front desk will be bypassed.
What a good deliverable contains
- A process map with owners, systems and volumes.
- A ranked opportunity list with assumptions and a confidence level.
- An owner for each recommendation, with a date.
- A cost model for build, run and internal time.
- An exception and control design for each automated step.
- A pilot plan with baseline metrics and a stop rule.
- A handover pack: access, credentials, runbooks and support contacts.
Metrics
Track cycle time, touch time, exception rate, rework rate, cost per transaction and adoption monthly. Measure the baseline before the pilot, or you will be arguing about impressions after launch. Choose one primary metric per process and one guardrail such as error rate or customer complaints.
The Bangladesh context
Local workflows often start on WhatsApp, move to a paper challan and end in an ERP. Mobile sales teams work from shared spreadsheets, and approvals happen by phone. That makes the assessment more valuable, not less: the biggest win is usually a single structured intake with a clear owner, not a complex AI model. Connectivity and power interruptions also belong in the design, with offline capture and a sync rule.
Where this fits in our work
If you want an assessment before committing to a platform, start with Business Automation. The pricing page explains how scope is built. For a related workflow example, read WhatsApp workflow automation, year by year.
Sources
- McKinsey & Company, The State of AI 2026, accessed 17 November 2026.
- NIST, AI Risk Management Framework, accessed 17 November 2026.
- ISO/IEC 42001:2023, Information technology — Artificial intelligence — Management system, accessed 17 November 2026.
- Object Management Group, Business Process Model and Notation (BPMN) 2.0.2, accessed 17 November 2026.
Responses (0)
No comments yet. Add the first one below.
Leave a response
Your email address will not be published. Required fields are marked *.