Dev48
Language
  • About
  • Services
  • Industries
  • Technologies
  • Articles
  • Contacts
Book a call
    Home/Articles/Thinking about automating a workflow 5 questions to ask first
Dev48
info@dev48.org+375 29 320-40-96WhatsApp@isky_by

© 2026 · All rights reserved.

Thinking about automating a workflow? 5 questions to ask first

Source: Nordcloud•November 17, 2026 at 12:02 AM

Thinking about automating a workflow? 5 questions to ask first

- The question isn’t what you can automate. It’s what is worth automating in the first place. Five questions can guide you toward the best approach. - Don’t automate inefficiency. First determine whether the work itself creates value or should be eliminated. - Fix friction at the source.…

- The question isn’t what you can automate. It’s what is worth automating in the first place. Five questions can guide you toward the best approach. - Don’t automate inefficiency. First determine whether the work itself creates value or should be eliminated. - Fix friction at the source. Redesign broken workflows before using technology to make them faster. - Apply AI where it creates real leverage. Prioritize processes with reliable data, clear rules and reusable capabilities. - Optimize for outcomes, not automation. Keep human judgment where it matters and measure success in business value, not simply work automated.

Automation has become an easy answer—“make it faster”—to a difficult question: “is the work worth doing?”

A process takes too long, so the instinct is to automate it. Employees spend too much time on repetitive work, so you introduce an AI agent. A workflow has too many handoffs, so you connect the systems. Each response assumes the same thing: that the work itself still makes sense. But if the work is unnecessary, inefficient or designed around constraints that no longer exist, automation may simply help the organization do the wrong things faster.

As AI moves deeper into business operations, research finds that AI-first CFOs are redesigning processes and workflows. They’re focused on embedding intelligence into how work gets done rather than simply layering AI onto legacy ways of working.

Khalid Siddiqui, Partner and Global Offering Leader for Finance Operations at IBM Consulting, has spent nearly three decades helping organizations redesign operations and intelligent workflows. For him, automation shouldn’t be the starting point. “First, understand the outcome you are trying to improve, the friction standing in its way and whether the process still makes sense for the business today. Only then can you determine where technology belongs.”

In other words, business leaders should ask, “What are we actually trying to change?” You can start by answering these 5 questions.

The first question isn’t how the work gets done. It’s what the work is meant to accomplish.

Siddiqui often sees organizations start with an activity they want to automate rather than an outcome they want to improve. But work only creates value in relation to the result it produces, which is why he recommends starting with a more basic question: why are we doing this work in the first place?

“Take invoice processing, for instance,” he says. “It may seem like an obvious automation opportunity. But the moment a team defines the challenge as “automate invoice processing, it has already shifted into solution mode.”

Siddiqui recommends starting with the outcome instead.

Are suppliers being paid accurately and on time? Are financial controls being maintained? Can those objectives be achieved with less manual intervention?

Siddiqui describes this as working backward from the outcome. Enterprise goals translate into business objectives, which translate into operational measures and transactional activities. The exercise forces a straightforward question: How exactly does this work contribute to the result the business ultimately cares about?

Without that connection, teams can spend significant effort optimizing activities without knowing whether they meaningfully influence business performance.

Siddiqui recalls a finance transformation effort at one IBM client that had identified 50 manual reconciliations as candidates for automation.

When the team examined them more closely, they discovered the reconciliations existed for very different reasons. Some stemmed from poor data quality, others reflected fragmented systems and disconnected handoffs, and only a portion represented genuine control requirements.

Automating all 50 would have accelerated the activity, but it wouldn’t have addressed the underlying causes.

“It would have been accentuating the inefficiencies,” says Siddiqui.

Therefore, the first question isn’t whether a task can be automated. It’s whether the task is necessary to achieve the desired outcome.

As a business leader, ask a simple question: If this work stopped tomorrow, what business outcome would suffer?

If the answer is little or nothing, you may have discovered something to eliminate rather than automate.

Once you understand the outcome and the work that genuinely supports it, the next step is identifying what’s getting in the way.

This is where visible problems can be misleading. A backlog, a long cycle time or employees spending hours answering requests often appear to be the problem. In reality, they may be symptoms of friction elsewhere in the workflow.

“You do not automate around friction,” says Siddiqui. “You should understand why friction exists.”

That friction can take many forms. It may originate in poor master data, inconsistent policies, fragmented systems, incomplete information, excessive approvals or too many handoffs between teams.

To understand what’s happening, Siddiqui recommends examining the end-to-end workflow rather than focusing only on the point where the problem becomes visible. Process-mining and task-mining tools can help identify bottlenecks, hotspots and unnecessary handoffs between people, systems and AI agents.

Siddiqui recalls working with a global building materials manufacturer whose collections team was spending nearly 40% of its time responding to approximately 1.2 million inquiries from sales and field teams rather than collecting cash. Resolving each inquiry could take 30–40 minutes.

It would have been easy to define this as an automation problem.

Instead, Siddiqui’s team asked a different question: Why were so many questions reaching the collections team in the first place?

The investigation revealed that many requests didn’t require a collector’s involvement at all. The necessary information already existed elsewhere in the process but remained difficult to access, fragmented across systems or trapped behind unnecessary approvals.

Only after understanding those root causes did technology enter the picture. The results were significant. Queries that once took 30–40 minutes could be resolved in roughly 3 minutes. Self-service adoption approached 90%. Day sales outstanding improved by approximately 2 days. The business also unlocked nearly USD 1 million in working capital.

The breakthrough came from reducing how much work required human intervention in the first place. “Don’t try to automate these exceptions,” Siddiqui says. “This is friction.”

Ask why the exception exists. The answer may point to a policy, process, data or systems issue that should be addressed upstream.

Few processes are designed to be complicated. Most become complicated over time.

A control gets added to address a risk. An approval is introduced to solve a problem. A new system creates another handoff. Responsibilities shift between teams. What begins as a series of sensible decisions gradually becomes a workflow shaped more by history than by intention.

Eventually, organizations can find themselves investing significant time and effort to maintain a process that no longer reflects how the business creates value.

That’s why Siddiqui believes leaders should distinguish between simplifying a process and redesigning it.

Simplification removes friction from an existing model. Redesign challenges whether the model itself still makes sense.

The difference matters because making a process more efficient is not always the same as making it more effective. When business priorities, operating models or customer expectations change, leaders may need to rethink how work is structured rather than simply improve how it flows.

The same principle applies beyond finance operations. In a recent IBM Think EMEA conversation, Prashant Jajodia, Managing Partner and EMEA Banking Leader at IBM Consulting, made the point directly: “The banks who are successfully deploying AI are fundamentally reimagining business processes.” Rather than layering AI onto existing processes, he says, they identify the processes that matter most to customers, employees and the bottom line, and reimagine them.

Siddiqui saw this firsthand through IBM’s Client Zero work in billing and invoicing. It involved more than 2,000 people and was delivering best-in-class metrics, although the process had accumulated layers of hierarchy, approvals, controls and policies over time.

Technology can make an existing process more efficient. But efficiency isn’t the same as relevance. If a process has been shaped by years of incremental decisions, leaders should first ask whether it still reflects how the business wants to operate today.

Siddiqui frames the question this way: “If you were designing this process today, with a blank sheet of paper, would you build it this way? If not, the opportunity may not be automation; it may be redesign.”

If so, only at this stage does it make sense to ask where technology can help. By this point, leaders should have identified the desired outcome, removed unnecessary work, addressed sources of friction and challenged the process design itself.

Now they can assess whether technology will create meaningful leverage.

Even then, readiness isn’t guaranteed.

A process may appear repetitive while still relying on inconsistent data. Different systems may contain conflicting information. Policies may vary by geography. Controls may be unclear. Exceptions may require more judgment than the process documentation suggests.

Technology introduced into that environment can increase complexity rather than reduce it.

Siddiqui has seen organizations deploy AI within workflows only to create additional interactions, reviews and handoffs around the technology itself.

They end up with “multiple handoffs with the AI agent rather than reducing them,” he says.

That’s why automation readiness depends on more than repetition.

The data must be dependable. The business rules should be understood. The workflow must be sufficiently consistent to support technology, and the value has to justify the investment.

Leaders should also think beyond automating individual tasks.

Historically, workflow automation often involved building standalone solutions for specific activities. Siddiqui believes there’s greater value in identifying the capabilities that can be reused across different workflows.

He compares a use case to a dish and the enabling capabilities to ingredients. Those ingredients might include extracting information, integrating systems, matching and validating data, applying AI models and completing transactions.

Just as ingredients can be used in many different recipes, those capabilities can support multiple workflows across the enterprise.

The economics look different when capabilities can be reused across workflows. Instead of treating each workflow as a separate technology investment, organizations can build reusable capabilities that create value across functions, business units and use cases.

That, however, still leaves an important question: Where should technology take over the work, and where does human judgment continue to matter?

Once leaders understand where technology can contribute, they still need to decide where human judgment remains essential to the outcome.

Siddiqui describes this as “targeted autonomy.”

In his experience, the greatest value often comes from being deliberate about where technology is applied. Some activities follow clear rules, rely on structured data and produce predictable outcomes. Others require interpretation, context and accountability.

“I’m a believer in targeted autonomy. Let AI run on its own where the task is well-defined and the stakes are clear, and keep humans in charge of the decisions that require judgment,” he says.

That distinction becomes particularly important in areas such as tax, audit and regulatory operations, where decisions can carry consequences beyond an individual transaction.

“Human in the loop is a must,” says Siddiqui.

Organizations therefore need to understand where human expertise has the greatest impact on the outcome.

When technology takes on routine and repeatable activities, people can focus on exceptions, judgment calls, risk assessments and decisions that require a deeper understanding of the business context.

A workflow that runs faster doesn’t necessarily create a better business outcome. Similarly, reducing the effort required to complete a task doesn’t automatically translate into meaningful business performance.

For Siddiqui, value should be evaluated through the lens of the outcome the organization is trying to achieve.

That value may appear in several ways. It may reduce operating costs, shorten cycle times, improve quality, strengthen compliance or reduce risk. In some cases, it may release capacity that can be redirected toward higher-value work. In others, it may improve measures that matter directly to the business, such as working capital, profitability, margin or customer experience.

No single metric tells the complete story. “For a business leader to demand what the value is, they cannot put it in one bucket,” Siddiqui says.

Understanding value requires looking across cost, speed, quality, risk and the business outcome itself.

Before automating, leaders should ask what outcome the work serves, what friction stands in the way, whether they’d build the process the same way today, whether it’s ready for technology and where human judgment belongs. Organizations that skip these questions speed up their inefficiencies. Those that ask them move beyond automating tasks to orchestrating people, agents, data and processes around the outcomes the business cares about, with value that compounds as each workflow is redesigned. The goal isn’t to automate more work. It’s to create more value from the work that remains.

Brand and Content Strategist

Thought Leadership

← All articles

What this article says

More in Cloud & Infrastructure

All →
2026 Gartner MQ Cloud Operational Databases Management Systems
Looker

2026 Gartner MQ Cloud Operational Databases Management Systems

Maximizing AI accelerator ROI with purpose-built storage
Looker

Maximizing AI accelerator ROI with purpose-built storage

Last year we built megakernels. This year we deleted them: The road to Together Compiler
Together AI

Last year we built megakernels. This year we deleted them: The road to Together Compiler

The Race to Remediation: Black October Showed that Threat Intelligence Isn’t Enough
JFrog

The Race to Remediation: Black October Showed that Threat Intelligence Isn’t Enough

Evaluate, observe, act: A practical approach to operating AI agents with confidence
Dynatrace

Evaluate, observe, act: A practical approach to operating AI agents with confidence

Claude Fable 5 vs 5.1 vs 5.2: What changes when you upgrade
SonarSource

Claude Fable 5 vs 5.1 vs 5.2: What changes when you upgrade

More from Nordcloud

Why good conversations have ‘pink noise,’ according to researchers
Nordcloud

Why good conversations have ‘pink noise,’ according to researchers

Your AI budget is a data definition problem: a data modernization strategy for CFOs
Nordcloud

Your AI budget is a data definition problem: a data modernization strategy for CFOs

Salesforce’s Paula Goldman on why AI needs a human at the helm | IBM
Nordcloud

Salesforce’s Paula Goldman on why AI needs a human at the helm | IBM

Anonymity until the last mile: 333 hijacked Tor exit nodes exposed
Nordcloud

Anonymity until the last mile: 333 hijacked Tor exit nodes exposed

The future of food delivery is here, and literally hundreds have tried it
Nordcloud

The future of food delivery is here, and literally hundreds have tried it

Streamhouse wants AI agents to know what’s true right now
Nordcloud

Streamhouse wants AI agents to know what’s true right now