Automate or Delegate: A Practical Framework for Accounting Firm Leaders to Choose Wisely 

AI in accounting has redefined what’s possible. 

More and more tasks can be handled by machines than ever before. 

Yet whether to automate or delegate remains a strategic crossroads—one that every accounting firm leader must learn to navigate to stay competitive. 

Having worked with 80+ firms as a process optimization consultant, I often help firm owners assess whether humans or machines are better suited for the job. 

In this article, you will learn: 

• When to let robots or humans do the job – and when to use both.
• Real-world case studies of accounting firm workflows faced with this choice.

• My 5-questions decision-making framework to help you decide for yourself.

Humans and Automation: A Partnership, Not a Battle

First off, this isn’t an either-or situation. 

Automation and delegation can – and should – coexist within accounting firms.  

To leverage the strengths of both, in fact, many developments are known for allowing humans and automation to collaborate

For example: 

• No-code automation platforms like Relay.app allow you to include a step for human approval or input, creating what’s called a human-in-the-loop automation.
• In the world of Robotic Robotic Process Automation (RPA) Process Automation (RPA), so-called attended robots have been known for doing what they can but calling for human help when faced with unknown situations.

• Also, we frequently suggest automated monitoring systems that flag instances that need further human review.  

How Much to Automate: 3 Real-World Case Studies

To illustrate the practical balance of automation and delegation, here are three real-world case studies from accounting firms we’ve helped. 

These examples highlight when automation made more sense and when human oversight was necessary, offering insight into how your firm can make similar decisions. 

Case Study 1: Reconciling Accounts and Flagging Variances (Automated Monitoring, Then Human Review)

An accounting firm sought us to streamline the process of identifying accounts that required reconciliation.  

We employed automation tools to extract the ending balances from bank statements and compare them with balances in their client’s accounting software (Xero). 

The former were read using AI – and the latter were extracted directly using Xero’s API.  

Then, the system automatically flagged accounts that had significant variances (e.g., differences greater than $100). 

Lastly, on a weekly basis, the automated system would alert a human accountant to review these flagged variances for further investigation or correction. 

Why Automation Worked:

In this instance, a discrepancy is an objective, mathematical variance – something automation can reliably handle. 

Given the data was accessible, the automation was able to handle the bulk of the repetitive data-scanning work while leaving human accountants to focus on reviewing the flagged discrepancies only. 

Case Study 2: Labeling Leads in the Inbox (Human Filtering, Then Automated Follow-Up)

At another accounting firm, the team wanted to label leads that came in via email or phone. 

The goal was simple: 

• Identify incoming inquiries from potential clients

• Automatically follow up with a scheduler link for the prospect to book a call

• Label and transfer this email to a different environment for the owner to prioritize and keep track of all revenue opportunities

Their phone operating system was already sending email summaries of phone inquiries. 

As such, we created an AI assistant that scanned incoming emails and categorized leads based on the email content. 

However, the automated assistant often misidentified existing clients or vendors as new leads, which caused confusion and required manual correction. 

We had to pull back. 

Ultimately, we allowed the owner’s (human) assistant to label the emails herself – and only then proceeded to automate the subsequent steps. 

Why Delegation Worked:

In this case, the task of labeling leads required a level of contextual information that we couldn’t easily provide the AI assistant. 

Could we have pulled a client list from another software tool (e.g., Quickbooks)? Or a list of existing vendors? 

Possibly—but even then, success wasn’t guaranteed. 

As a result, we decided to let humans handle the first step since they could quickly and accurately distinguish between potential clients, existing clients, or vendors. 

Case Study 3: Determining 1099 Eligibility (AI Assessment, Then Human Confirmation)

Another US-based accounting firm wanted to streamline the process of identifying which vendors required IRS Forms 1099-NEC or 1099-MISC based on their payments during the year. The task was repetitive and time-consuming, but it also required some judgment to ensure accuracy. 

We implemented an AI assistant that pulled vendor payment records and included the most up-to-date IRS guidelines for 1099 eligibility. For example, the AI knew that vendors who received over $600 in nonemployee compensation were eligible for 1099-NEC, and those with rent or royalty payments for 1099-MISC. 

The AI systematically assessed each vendor based on key criteria, such as payment thresholds, business types, and exclusion rules. It was also prompted to provide a confidence score (e.g., 90%) and a clear reasoning for its determinations. 

Why AI Worked:

Although the AI-generated determination wasn’t 100% accurate, this assistant greatly reduced the amount of time accountants needed to spend manually reviewing each vendor’s eligibility. 

When a vendor’s payment category didn’t fit neatly into the stated rules or the AI couldn’t confidently determine eligibility, human oversight was still available to ensure accuracy and correct any errors the AI may have missed. 

Factors to Consider When Deciding to Automate or Delegate

As you can see, several factors made us either scale the automation back (and use humans) or scale humans back (and use more automation or AI). 

I’ve narrowed down the factors into 6 questions you can ask to make this assessment easy for yourself. 

These will help you strike the right balance between efficiency and human oversight, ensuring optimal outcomes for your accounting firm. 

1. Insufficient or Inaccessible Information

Question: Is the information needed not easily accessible?

When the data the automation needs is challenging to access (e.g., institutional knowledge not documented), automation may lead to errors. Human judgment can fill in gaps and navigate uncertainty. 

Example:

In Case Study 2, the task of labeling leads in an inbox required knowing existing clients and vendors, which the automation tool didn’t have access to. In this instance, delegating this first task to a human ensured higher accuracy.

2. Highly Variable vs. Rule-Based Process

Question: Is the process subject to frequent changes or high variability?

Automation struggles with constantly evolving or highly variable tasks because it requires reprogramming to adapt. Humans, on the other hand, can adjust to changes on the fly. 

That said, there are a lot of rule-based decisions and processes in accounting—more than we often like to admit. 

Example:

In Case Study 3, determining 1099 eligibility involved vendors with varied circumstances. While the AI handled the initial review, human accountants were necessary to review their work.

3. High-Stakes Tasks

Question: Does the task involve significant consequences if handled incorrectly?

Tasks with high stakes, such as financial impact or regulatory compliance, should be handled by humans to minimize risks. Automation can assist, but humans should have the last call. 

Example:

In Case Study 1, reconciling accounts incorrectly can have financial implications. While automation flagged variances, human accountants made the final review. 

4. Relationship-Based Communications

Question: Does the task involve direct communication or relationship-building with clients?

Tasks requiring personal interaction or relationship management should be delegated. Automation lacks the emotional intelligence and context required for meaningful client communications 

I may change my mind, but I want clients to feel bonded with our team. It’s healthy. 

Example:

In Case Study 2, automating the process of labeling leads caused miscommunication with existing clients and vendors. Delegating the task to a human ensured that client relationships were managed appropriately and confusion was avoided. 

5. Low Frequency

Question: Is the task too infrequent for automation to be worthwhile?

Automating infrequent tasks can be more trouble than it’s worth. If a task doesn’t occur often, the setup and maintenance required for automation might not justify the time saved. 

Example:

For obvious reasons, if the process only applies to one small client or it happens once in a blue moon, it’s just not worth it.

Automation Can Take On More Than You Think

Automation offers more possibilities than most realize. 

While both humans and technology play roles in accounting operations, many underestimate what automation can achieve. 

Use the 5-question analysis above to identify opportunities to streamline your firm and start enabling your team to delegate more tasks to technology. Because if you don’t, someone else will. 

About the Writer

---