A poor workflow design can generate more cost than the process it replaces

There is one number that circulates in all reports on business automation that no software vendor tires of citing: the three-year ROI of 248 percent that Forrester calculated for Power Automate. [1]

It is a real fact, with a published methodology and a verifiable sample. And it's fully compatible with poorly planned automation destroying value rather than creating it.

The difference between the two scenarios isn't in the tool. It's what happens before you open the flow editor.

Gartner identifies three recurring patterns in automation projects that fail in a medium-sized company: automating the symptom instead of the root process, implementing flows without a clear owner within the organization, and designing without considering exceptions. [6]

The three have one thing in common: they are pre-implementation design errors, not technical platform problems. And all three come at a real cost. Deloitte estimates that 30% of organizations that failed the automation pilot phase lost between 50,000 and 200,000 euros in abandoned projects. [2]

The three most common design errors (and what causes them in practice)

Mistake 1. Automate the symptom, not the process

‍The most common symptom in medium-sized companies is the accumulation of coordination emails: approvals that circulate by email, manual notifications about order status, reminders sent one by one. The intuitive reaction is to automate those emails: create a workflow that automatically sends them when a condition is met.

The problem is that these emails are the symptom of a process that is not properly structured. If the invoice approval process requires three people to respond to a chain email, automating the emails doesn't solve the process: it makes it faster under normal conditions and more opaque when it fails. 77% of medium-sized companies do not have their approval workflow process digitized. [5]

Automating communication about a non-digitized process produces, in the best case, surface efficiency.

The detection is simple: if the workflow design requires more than two exceptions (“if the amount exceeds X...”, “if the person responsible is on vacation...”), it is likely that the symptom is being automated. The root process needs to be redesigned before it can be automated.

Mistake 2. Ownerless flows

‍A workflow in production is software in production. It has external dependencies (third-party APIs that change their parameters), input data that varies (file formats that evolve) and business conditions that change over time.

When something fails at two o'clock in the afternoon on a Tuesday and the process that that workflow supports stops, someone has to be responsible for fixing it.

In most mid-size enterprise automation implementations, that someone isn't defined. The workflow was built by someone who had technical capacity at the time — an external consultant, the IT technician or the most knowledgeable employee in the department — and, when that person is gone, no one knows how to intervene.

McKinsey estimates that 62% of the gap between automation potential and actual automation is explained by deficiencies in process design, not technology. [8] Fuzzy ownership is the first of these deficiencies.

The minimum standard for any workflow in production is simple: an identified person who is alerted when the flow fails, who can access the error logs, and who has the authority to decide if the flow stops or continues with a manual exception while the problem is being diagnosed.

Without that defined person before the go-live, the workflow is a technical debt with an unknown expiration date.

Mistake 3. Design without exception management

‍The happy flow—the path the process takes when everything works as expected—is always the easiest to automate. The problem is that real processes rarely follow that path 100% of the time.

A vendor sends an invoice in a different format, an external API returns a 503 error, a required field arrives empty, or a record doesn't exist in the CRM.

Workflows without exception management fail in two ways: silently or noisily. Either the data is poorly processed without anyone noticing it—generating errors that are discovered weeks later—or the flow stops completely blocking the process.

The case documented by Forrester in the pharmaceutical industry—11,000 hours saved with 72 automations—identifies as a key factor that each automation had a defined fault response SLA before deployment. [7]

A practical rule for evaluating the maturity of a workflow design before implementing it: if the flow is not defined as to what happens when each of its steps fails, it is not designed. It's sketched.

The difference between a sketch and a design is the cost of maintenance in production.

The map of the invisible cost in poorly planned automation

The cost of poorly designed automation rarely appears on a single invoice. It is distributed over time in items that are not always attributed to automation.

‍

When automation does deliver the promised ROI

The 248% ROI that Forrester documents for Power Automate is not an exceptional case. Deloitte reports that organizations that passed the pilot phase of automation—with appropriate design, defined ownership, and exception management—achieved a 32% reduction in operating costs. [2]

Musixmatch saved 47 days of engineering work in four months with n8n automations. [9] McKinsey estimates that 57% of working hours are automatable with existing technology today. [3]

These results have one thing in common: they all start from a Audit of the process before implementation. Identify precisely how much time the process consumes today, what percentage of that time is automated repetitive work, what are the common exceptions and who will be the owner of the flow once in production.

Without that information, any ROI estimate is speculative.

The three indicators that an automation is ready to be implemented

The process is documented step by step, including exceptions

‍Not the ideal flow, but the real flow as it happens today, with all its variants. If it can't be documented, it can't be reliably automated.

There is a person identified as the owner of the automated process

‍With the ability to access logs, with an alert channel when the flow fails and with the authority to stop it if necessary.

The flow design includes what happens at each possible point of failure.

‍Not just the happy path. Each node in the flow has its response to an error defined: retry, notification, escalation or stop.

The most powerful automation tool on the market produces technical debt if the process it automates was not well defined.

The correct sequence is not: choose tool → automate → optimize.
It is: document process → identify owner → design flow with exceptions → choose tool → implement.

Process diagnosis is part of the service, not an expendable previous step

64% of ERP projects go over budget. [10] Panorama Consulting's analysis of the causes identifies the lack of definition of scope as the main factor, over and above technical implementation problems.

Process automation replicates the same pattern on a smaller scale: implementations that fail don't fail because of the tool, they fail because the process was not sufficiently defined when the flow was built.

This has a direct implication for any company that is evaluating an automation initiative: the value of a consultancy that spends time mapping the process before opening the flow editor is not an additional cost. It's the difference between building something that works in production and building something that works in the demo.

65% of ERP users consider it difficult to access their own data within the system. [11] Automating the extraction of data from a misconfigured ERP does not solve the problem of accessing the data: it hides it behind a flow that seems to work.

The right data in the right system, with the right process, in the hands of the right person: that's what generates the 248% ROI. [7] The workflow is the last step, not the first.

References

1. Forrester Research. (2024). The Total Economic Impact™ Of Microsoft PowerAutomate. Forrester Consulting, commissioned by Microsoft. https://tei.forrester.com/go/microsoft/powerautomatetei/index.html —ROI of 248% over three years on a composite organization of 5,000 employees; NPV $39.85 M; total benefits $55.93 M; costs $16.08 million; payback period of less than six months. The report itself details that these benefits are achieved when automated processes are well defined and have a clear owner.

2. Deloitte. (2023). Automation with Intelligence: 2022 Global Automation Survey.Deloitte Insights. https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2022-survey-results.html — Organizations that passed the automation pilot phase achieved a 32% reduction in operating costs. 30% of those who failed the pilot reported losses between $50,000 and $200,000 in abandoned projects.

3. McKinsey & Company. (2025). A new future of work: The race to deploy AI and raise skills in Europe and beyond (November 2025). McKinsey Global Institute. The report updates the estimate of working hours that can be automated with existing technology: from 30% (2023 estimate) to 57% in 2025. Employees waste approximately 200 hours a year on tasks that could be automated with current tools.

4. Gartner. (2024). Predicts 2026: Low-Code and No-Code Application Platforms.Gartner Research. Gartner projects that by 2026, 75% of new business applications will be built with low-code or no-code platforms. The same projection warns that 60% of poorly planned automation projects will generate technical debt that requires external intervention in the following 18 months.

5. ERP News. (2024). Electronic Workflow Process Gaps Kill an Estimated 20% of ERP Productivity. ERP News. 77% of medium-sized companies do not use the workflow management features of their ERP. Organizations lose between 15% and 20% of the potential value of their ERP due to non-digitized approval and coordination processes.

6. Gartner. (2023). How to Build a Business Case for Process Automation. Gartner Research (ID: G00797865). Gartner identifies three recurring patterns of failure in automation projects: (1) automating the symptom instead of the root process, (2) lack of clear ownership of the automated process, and (3) lack of exception management in the initial design.

7. Forrester Research. (2024). The Forrester Wave™: Robotic Process Automation, Q1 2024. Forrester Research. The report analyzes cases of automation in companies with between 500 and 5,000 employees. In the pharmaceutical sector, an anonymous company saved 11,000 hours a year with 72 properly designed Power Automate automations. The cited key: each automation had an identified owner process and a defined fault response SLA.

8. McKinsey & Company. (2024). The state of AI in early 2024: Gen AI adoptionspikes and other trends. McKinsey Global Institute. 45% of the time spent on approval and coordination processes in medium-sized companies could be automated with existing technology. The gap between automation potential and real automation is 62% explained by deficiencies in process design, not technology.

9. Music match. (2025). How Musixmatch saved 47 engineering days in 4 months withn8n. n8n Blog. https://blog.n8n.io/musixmatch/ — Real case published by n8n: Musixmatch implemented data synchronization automations and pipeline notifications with n8n, eliminating repetitive manual work from the engineering team. The savings of 47 engineering days in 4 months were calculated on the previous time spent on manual tasks identified in the pre-implementation audit.

10. PanoramaConsulting Group. (2024). 2024 ERP Report. Panorama Consulting. https://www.panorama-consulting.com/resource-center/erp-report/ —64% of ERP projects exceed the initial budget. 75% exceed the original deadline. The most frequently cited factor for failure is the lack of definition of scope during the design phases, not technical implementation problems.

11. SaaSworthy/ multiple compiled sources. (2024). Top 50 ERP Statistics That Will Define 2025. https://www.saasworthy.com/blog/top-erp-statistics — 65% of users consider that accessing their own data within the ERP is difficult. Only 11% of companies believe that their ERP captures all the non-financial information needed to monitor operational KPIs.