The France Num survey published on September 28, 2026, reports that 40% of French microbusinesses and SMEs now use artificial intelligence, while only 11% use it to automate tasks. The gap is a useful reminder: adopting a tool is not the same as integrating it into a real process and funding its entire lifecycle, as detailed in the 2026 France Num survey.
The starting point: obtain a production-ready budget
Consider an illustrative scenario: a department receives requests by email, checks information in a CRM, creates a transaction in the ERP, and informs the requester. The objective is not to obtain a quick quote for “automating emails.” It is to produce a budget in euros that covers design, interfaces, licenses, testing, operations, and corrective maintenance.
Management should receive a one-page decision document containing low, central, and high estimates, the cost per successful transaction, expected monthly net value, and break-even expressed in months. This makes the project comparable with other investments instead of reducing it to a software subscription.
Our articles on Power Automate licensing costs in 2026 and Make’s AI credit calculation explain platform-specific billing. This playbook addresses a different issue: calculating the full cost of the automated process, regardless of the selected technology.
Prerequisites
Before requesting a quote, gather:
- A process owner: the person accountable for the business outcome and authorised to decide rules, exceptions, and human controls.
- A map of the current process: activities, roles, systems, decisions, exchanged data, and exception paths. If it does not exist, use our process-mapping playbook and document it in BPMN Studio.
- Verifiable operating data: application logs, request history, actual handling time, manual rework, errors, and waiting time.
- Loaded internal labour costs: finance must supply the employer’s full cost, not employees’ net salaries.
- Technical access information: API documentation, test accounts, authentication rules, connector limitations, and contracts for the systems being connected.
- A decision rule: maximum budget, acceptable payback period, required service level, and the business impact of downtime.
Do not select Power Automate, n8n, Make, RPA, or custom development yet. Choosing a tool before documenting the constraints can move costs into premium connectors, additional code, or long-term maintenance.
The step-by-step process
1. Document the process as it actually operates
Who does what: the process owner walks through a real transaction with operational users. The analyst records every entry, check, delay, handoff, and correction.
Realistic duration: one focused workshop followed by asynchronous validation.
Deliverable: an as-is BPMN diagram and a matrix listing each activity’s owner, application, input, output, and exception paths.
The map must include spreadsheets, temporary files, and workarounds rather than describing the official procedure alone. D1 Consulting’s business process audit and mapping service establishes this baseline before any technology choice is made.
2. Measure the manual baseline
Who does what: the operational manager records active time, waiting time, and rework. Finance supplies loaded hourly costs. The analyst consolidates volumes and separates standard cases from exceptions.
Realistic duration: a representative operating cycle, supplemented by available historical data.
Deliverable: a baseline sheet containing volumes, manual time by case type, rework rate, and the process’s current cost.
Use the following formula:
annual baseline cost = annual volume × average manual time × loaded hourly cost
Add measurable costs caused by errors, including credit notes, duplicate records, re-entry, complaints, or penalties. Do not insert an assumed productivity percentage. Use observed data or assumptions that are explicitly labelled as such.
3. Design the target process and its exceptions
Who does what: the process owner decides which activities remain human. The integrator defines triggers, rules, connectors, and recovery mechanisms. IT validates access, security, and operational requirements.
Realistic duration: a short design sprint, extended when business rules remain contradictory.
Deliverable: a target BPMN diagram, an exception matrix, and a list of required human controls.
Separate deterministic rules from tasks requiring interpretation. For example, checking whether an identifier is present does not require AI. Add an AI component only where conventional rules cannot adequately perform the task.
Microsoft similarly structures automation around discovery, design, build, testing, deployment, and operations in its Holistic Enterprise Automation Techniques guidance.
When personal data is exchanged, document purpose, data minimisation, permissions, and log-retention rules. The French data protection authority recommends limiting API exchanges to strictly necessary data and securing access keys in its API security guidance.
4. Estimate the whole build, not only the workflow
Who does what: the integrator divides the solution into estimable work packages. IT confirms environment and operating requirements. Business users estimate their contribution to workshops, tests, and documentation.
Realistic duration: an estimation phase carried out after access, connectors, and sample data have been reviewed.
Deliverable: a detailed estimate showing assumptions, exclusions, and ownership.
The initial budget must include:
| Cost item | What to estimate |
|---|---|
| Discovery | Mapping, rules, exceptions, and metrics |
| Build | Workflows, scripts, forms, and optional AI rules |
| Integration | APIs, connectors, webhooks, gateways, and service accounts |
| Data | Cleaning, mapping, migration, and controls |
| Security | Secrets, permissions, logging, and compliance |
| Testing | Standard paths, failures, duplicates, and outages |
| Deployment | Environments, production release, and rollback |
| Adoption | User procedures, administration, and support |
Require a separate line for every unverified assumption. An API assumed to exist, unknown data quality, or an untested connector should not disappear into a fixed-price total.
5. Calculate recurring cost using the correct billing units
Who does what: the integrator simulates executions. Procurement validates licensing. IT estimates hosting, monitoring, backups, and support.
Realistic duration: a simulation cycle using observed volumes and peak-load scenarios.
Deliverable: a monthly operating budget with a volume-based adjustment formula.
As of October 6, 2026, the platforms use different billing units:
| Platform | Unit to monitor | Official pricing reference |
|---|---|---|
| Power Automate | User, bot, or hosted machine | Premium at €13 per user/month, Process at €130 per bot/month, and Hosted Process at €186.30 per bot/month on Microsoft’s pricing page |
| n8n | Complete workflow execution | Starter at €20 per month, Pro at €50, and Business at €667 when billed annually on the n8n pricing page |
| Make | Credits consumed by actions | At 10,000 monthly credits, Core at $12, Pro at $21, and Teams at $38 on the Make pricing page |
These prices are not the project budget. Add hosting, paid API calls, AI consumption, logging, monitoring, corrective maintenance, and incident handling. Also include the future cost of changes to the CRM, ERP, authentication method, or data format.
6. Build the decision scenarios
Who does what: finance consolidates costs. The process owner validates which benefits are genuinely recoverable. Management decides according to risk and investment capacity.
Realistic duration: one decision meeting after the assumptions have been reviewed by both business and technical owners.
Deliverable: a go-or-no-go paper containing a cautious scenario, a central scenario, a budget ceiling, and stopping conditions.
Use these formulas:
monthly net value = avoided manual cost – residual human cost – monthly operating cost
break-even in months = initial investment ÷ monthly net value
If released time does not directly reduce expenditure, state how it will be used: processing additional transactions, shortening lead times, reducing risk, or clearing a backlog. Our guide to measuring Power Automate ROI explains how to continue this analysis after deployment.
How to measure whether it works
- Cost per successful transaction: recurring costs plus amortised initial investment, divided by transactions completed without rework.
- Net time released: baseline manual time minus residual manual work, supervision, and incident handling.
- Exception rate: transactions leaving the standard path divided by total volume, classified by root cause.
- Actual break-even: the date on which cumulative net benefits exceed cumulative costs, compared monthly with the approved forecast.
Technical logs should feed these indicators automatically. Microsoft notes that flows are not “set and forget” solutions and recommends regular monitoring of runs, errors, usage, and connectors in its flow-monitoring guidance.
Mistakes that cause the project to fail
- Buying licenses before mapping the process: the business requirement may be distorted to fit the tool’s limitations.
- Accepting a build-only quote: require separate costs for integration, testing, deployment, monitoring, and maintenance.
- Ignoring internal staff time: workshops, data preparation, testing, and approval consume resources that belong in the budget.
- Testing only the standard path: missing data, duplicates, rejection, outages, and recovery must be included. Microsoft recommends testing every possible pattern and outcome, followed by a final user check, in its Power Automate testing strategy.
- Automating an unstable process: every unresolved rule change becomes a technical correction.
- Confusing a demonstration with production readiness: a demo proves neither error handling nor rollback. Our article on functional acceptance testing before sign-off lists the evidence to require.
- Leaving the automation without an owner: every workflow needs a business owner, a technical owner, and an escalation procedure.
Getting support
A business leader reading this article can ask us to perform a process audit and mapping engagement, followed by budgeting and implementation through our Automation & Process Optimisation service.
D1 Consulting handles BPMN Studio modelling, baseline measurement, target-process design, lifecycle-cost calculation, architecture selection, implementation, acceptance testing, and operating dashboards. Depending on your environment, we can deploy the solution through Power Automate and Microsoft 365 or build n8n and Make workflows.
👉 Book your free 30-minute diagnostic to identify the budget structure and operational data required for your first process.
