Useful planning ranges—not fixed prices
Two projects called “automation” can differ by an order of magnitude. Use ranges to decide whether a business case deserves discovery, then price the defined scope, systems, controls, and delivery conditions.
| Scope | Indicative planning range | Typical boundary |
|---|---|---|
| Configuration or simple integration | Low four figures | One or two supported tools, stable rules, limited variation |
| Focused custom workflow | Five figures | Several steps or systems, user review, monitoring, real exceptions |
| Multi-system operational solution | High five to six figures or more | Complex rules, migration, roles, critical integrations, broader rollout |
Eight factors that change automation cost
- 01
Process variation
More order types, document layouts, regions, policies, and exceptions require more design and testing.
- 02
System integration
Supported APIs are usually simpler than legacy databases, file exchanges, or interfaces with limited documentation.
- 03
Data readiness
Duplicates, missing identifiers, conflicting sources, and unclear ownership create cleanup and reconciliation work.
- 04
Roles and controls
Permissions, approvals, segregation of duties, audit history, and sensitive data add important scope.
- 05
User interface
A background transfer costs less than a review queue, operational dashboard, portal, or mobile workflow.
- 06
Reliability requirement
Retries, monitoring, high availability, recovery, and support depend on how costly interruption would be.
- 07
Migration and rollout
Historical data, parallel running, training, region-by-region launch, and change management require real effort.
- 08
AI evaluation
Variable documents or language require representative test data, quality thresholds, review design, and ongoing monitoring.
Budget the full lifecycle
- Process discovery, measurement, and target design.
- Product configuration, integration, or custom development.
- Data cleanup, mapping, and migration.
- Security, privacy, access, and architecture review.
- Testing with representative normal and exception cases.
- Training, rollout, documentation, and manual fallback.
- Cloud, software, model, messaging, and API usage.
- Monitoring, support, maintenance, vendor changes, and improvements.
How to estimate a first project responsibly
- 01
Bound the process
Name the trigger, completion event, users, systems, normal path, and exceptions included.
- 02
Bring real examples
Use representative records, documents, messages, and failures to test assumptions.
- 03
Confirm access
Check API, export, authentication, permission, and test-environment constraints early.
- 04
Separate must-have from later
Protect the business result while keeping the first delivery small enough to learn.
- 05
Price uncertainty explicitly
Use discovery, options, assumptions, and contingencies instead of hiding unknowns in one fixed number.
Compare the project with the cost of continuing
Measure monthly handling time, correction work, waiting, external fees, overtime, delayed revenue, and the growth that the current process cannot absorb. Avoid treating every saved hour as cash; state how the released capacity will be used.
| Decision measure | Calculation |
|---|---|
| Current monthly handling | Runs × minutes per run ÷ 60 |
| Current monthly labour cost | Handling hours × loaded hourly cost |
| Expected monthly saving | Current cost × conservative improvement rate |
| Simple payback | Implementation cost ÷ monthly saving |
| First-year net value | 12-month saving − implementation − recurring cost |
Reduce cost without weakening the outcome
- Automate one process or order type before multiple departments.
- Use existing product capabilities where they meet the important requirement.
- Keep rare, judgement-heavy exceptions manual in the first release.
- Integrate only the systems needed for the target result.
- Clean the minimum data required instead of launching a broad data programme.
- Reuse identity, hosting, monitoring, and design patterns already approved by the business.
- Define acceptance measures early to prevent endless subjective refinement.