SJKConsult

When should you automate a business workflow?

Not every manual process needs automating. How to find the workflows that pay back, and avoid the ones that don’t.

SJK ConsultWritten by the engineers who build the work
15 JAN 2025 · 7 MINin
ON THIS PAGE · 7 SECTIONS▾
01The 3 Rules for Automation ROI02Red Flags: When NOT to Automate03The Best First Automation Projects04How We Approach Automation Projects05Real Example: Support Ticket Routing06Getting Started07Need Help?

Automation sounds great in theory, but not every manual process is worth automating. We've seen companies waste thousands on automating the wrong things. Here's how to avoid that.

01#The 3 Rules for Automation ROI

Rule 1: Volume × Time = Opportunity

Calculate how much time your team spends on a task per week. If it's less than 2 hours total, automation probably isn't worth it yet.

Example:

  • Task takes 10 minutes
  • Done 3 times per week
  • Total: 30 minutes/week = Not worth automating (yet)

Vs.

  • Task takes 10 minutes
  • Done 50 times per week
  • Total: 8+ hours/week = Strong automation candidate

Rule 2: The "Rule-Based" Test

If you can write down clear if/then rules for a process, it's probably automatable. If it requires human judgment on every instance, it's not ready.

Good automation candidates:

  • "If customer email contains 'refund', forward to finance team"
  • "When invoice is paid, update CRM and send thank you email"
  • "Every Monday at 9am, pull last week's sales data into report"

Bad automation candidates:

  • "Read this proposal and decide if we should work with them"
  • "Look at these designs and pick the best one"
  • "Write a unique response to each customer based on context"

Rule 3: Stability Before Automation

Don't automate a process that's still changing. If you're still figuring out the best way to do something, automation will just lock in a bad process.

Wait until:

  • The process has been stable for 3+ months
  • Your team agrees this is "the way" to do it
  • Changes are rare (not weekly tweaks)

02#Red Flags: When NOT to Automate

  1. The process takes 5 minutes once a month
    Just do it manually. The ROI isn't there.
  2. You're automating to avoid fixing a broken process
    Fix the process first, then automate.
  3. The cost of errors is very high
    If a mistake could lose a major customer or cause compliance issues, keep humans in the loop.
  4. The tool landscape is changing rapidly
    If you're switching between tools every 6 months, wait until things stabilize.

03#The Best First Automation Projects

Start with these high-impact, low-risk workflows:

1. Data Entry Between Systems

Moving data from one tool to another (CRM to spreadsheet, support ticket to project management, etc.)

Why it works:

  • High volume
  • Rule-based
  • Low error risk
  • Immediate time savings

2. Notification & Alert Routing

Sending alerts, notifications, or tasks to the right people based on triggers.

Why it works:

  • Fast ROI
  • Easy to test and refine
  • Reduces response times
  • Improves customer satisfaction

3. Report Generation

Pulling data and creating recurring reports.

Why it works:

  • Very repeatable
  • Time-consuming when manual
  • High visibility (leadership notices)
  • Clear success metrics

04#How We Approach Automation Projects

At SJK, we follow a simple framework:

  1. Map the current state (1 week)
    Watch your team work. Document every step. Find bottlenecks.
  2. Calculate ROI (1 day)
    Time saved × hourly cost × 12 months vs. automation cost.
  3. Build MVP automation (2-4 weeks)
    Start with the highest-value portion. Get something working fast.
  4. Test & refine (2 weeks)
    Run in parallel with manual process. Catch edge cases.
  5. Scale (ongoing)
    Once proven, expand to other workflows.

05#Real Example: Support Ticket Routing

Before automation:

  • 50 tickets/day manually reviewed and assigned
  • 15 minutes per ticket = 12.5 hours/day
  • Delays in response time
  • Inconsistent routing decisions

After automation:

  • 80% of tickets automatically routed based on keywords and customer data
  • Only complex cases reviewed manually (10 tickets/day)
  • 2.5 hours/day of manual work
  • Saved 10 hours/day = $50K+/year in labor costs

Cost to build: $15K
Payback period: 4 months

06#Getting Started

Ask yourself:

  1. What manual task does your team complain about most?
  2. How many hours per week does it take?
  3. Can you write down clear rules for how it should work?

If you answered #1 and #2, and the answer to #3 is yes, you've found a great automation candidate.

07#Need Help?

We help teams identify and automate their highest-ROI workflows. Book a free 20-minute call to discuss your specific situation.

Was this useful?
WRITTEN BYSJK ConsultBetween us: 20+ years of combined experience at Amazon, Uber, Expedia and Microsoft, and degrees from BITS Pilani and Delhi College of Engineering. The people who write these notes answer the briefs.Got this problem? Send us a two-line brief →

Keep reading

All insights →
ENGINEERING NOTE

The exception queue nobody owned

Automations rarely fail on the work they were built for. They fail on the few items that don't fit. Give exceptions an owner and a deadline first.7 MIN · OCT 2026 · KESHAV YADAV
ENGINEERING NOTE

When not to use AI

If a rule can do it and someone has to explain the result, use the rule. A practical test for where AI helps and where it gets in the way.7 MIN · OCT 2026 · KESHAV YADAV
ENGINEERING NOTE

Why your first automation should be boring

Pick the dull, high-volume task first. Boring wins compound, and they buy the trust you need for the interesting ones.6 MIN · OCT 2026 · KESHAV YADAV
ENJOYED THIS? · ONE LIKE IT EVERY MONTH

One practical email a month. No filler.

SJK Engineering Notes: a lesson, an annotated artifact and a checklist, from work we actually shipped.

Monthly. Unsubscribe in one click. Privacy notice applies.