Knowing which tasks to automate is far more important than simply deciding to use AI. Every business owner hears the same advice: “Automate your business.” “Use AI to save time.” Yet almost nobody explains where to begin, and that’s exactly why so many automation projects never get off the ground.
Instead, founders often tackle the most complicated workflows first—judgement-heavy decisions, complex client processes, or large operational systems—and become frustrated when the automation falls short. Others don’t start at all because the sheer number of recurring tasks feels overwhelming. Meanwhile, the repetitive admin continues to consume hours every week.
This post gives you a practical way forward. You’ll learn a simple framework for identifying the highest-value opportunities, a scoring system to prioritise them objectively, and a proven sequence that builds momentum and recovers meaningful bandwidth within weeks instead of months.
Key Takeaways
- Successful automation starts with choosing the right tasks, not trying to automate everything at once.
- Every recurring task falls into one of four categories: fully automatable, heavily assisted, supervised, or human-only.
- Prioritise automation using a simple scoring system based on frequency, time cost, and automation fit to identify the highest-impact opportunities.
- Category 1 tasks, such as lead response, reporting, and routine notifications, typically deliver the quickest and most reliable wins.
- Automation should remove repetitive work while allowing people to focus on decisions, relationships, and strategic thinking.
- Common mistakes include automating complex tasks first, trying to transform the whole business at once, and confusing automation with replacing people.
- Automation is most effective when built on a foundation of business context, connected data, and AI-driven intelligence rather than isolated tools.
- Treat automation as an ongoing discipline by regularly auditing, scoring, and improving recurring tasks instead of approaching it as a one-time project.
Why “Just Automate Your Business” Is Useless Advice
The problem with most automation advice is that it is generic. It treats automation like a single decision: yes or no, automate or do not. In reality, every recurring task in your business sits on a spectrum. Some are fully automatable today. Some need human judgement for the final 5%. Some require so much context that automating them would take longer than doing them yourself.
If you try to treat them all the same, you will fail at the ones that need judgement, over-engineer the ones that do not, and miss the obvious wins sitting right in front of you.
The second problem is sequencing. Most founders try to automate the biggest, hardest, most visible task first. The one that has been annoying them for years. Usually a complex client-facing workflow. It feels like the right target because it is the loudest pain. But it is almost always the wrong place to start. The psychology of automation is that early wins compound. You need to feel the system handling something before you trust it with something bigger.
Which means the answer to “which tasks to automate first” is almost never “the biggest one.” It is “the one that gives you the fastest, cleanest win so you build momentum.”
Third problem: most founders have never actually mapped their recurring tasks. They work from memory. And memory is wildly inaccurate when it comes to where your time goes. Ask someone how many hours they spent on admin last week, and they will undercount by 50%. Until you see the list on a page, you cannot make good decisions about which tasks to automate.
The Four-Category Scoring Framework
Here is the framework I use with every client. Every recurring task in your business fits into one of four categories. The category determines what you do with it.
Category 1: Fully Automatable
These are tasks where a well-configured system can do the job from start to finish without human input. Input comes in, output goes out, no one needs to touch it.
Examples:
– Sending a welcome email when a new lead fills out a form
– Scraping a list of new business registrations each morning and adding them to a CRM
– Generating a weekly revenue report from your data sources
– Responding to missed calls with a text message asking if they want a callback
– Moving an opportunity to the next pipeline stage when a specific condition is met
The test: if you described this task to a careful assistant and gave them clear rules, could they do it without ever escalating? If yes, it is fully automatable. These are your biggest wins. They cross off permanently. They compound. They are where you start.
Category 2: Heavily Assisted
The AI does 80% of the work. You steer the last 20%. The task still happens without you initiating it, but you review or tweak the output before it goes live.
Examples:
– Drafting replies to inbound enquiries (AI writes, you approve)
– Generating a first draft of a proposal from a client brief
– Summarising meeting notes and flagging action items for review
– Writing social media captions for a set of pre-approved topics
– Drafting follow-up sequences to warm leads
These are where most founders get the biggest bandwidth recovery per hour invested. You are not removing yourself from the task; you are removing the heavy lifting. The thinking, the writing, the assembly. You show up at the end, make a judgement call, and ship.
Category 3: Supervised
The AI does the work. You review it, but only in aggregate. You do not touch each instance; you spot-check the system’s output and intervene if something drifts.
Examples:
– An AI receptionist handling inbound calls, qualifying leads, and booking appointments
– A database reactivation sequence running to 319 dormant contacts (the James case study: $49,000 recovered without the team touching a single message)
– A lead response system contacting every new enquiry within 90 seconds across SMS, email, and voice
– A daily brief that synthesises meetings, messages, and data into a morning summary
Supervised tasks are where the system starts to feel like a team member. You are not approving each interaction; you are checking the results over a week and adjusting the instructions when something is off. This is what most people imagine when they think “AI running the business.”
Category 4: Human-Only
These are tasks where the judgement, relationship, or creativity required is beyond what you would trust to a system, even with heavy supervision. They stay with a human.
Examples:
– Firing a team member
– Pricing a major strategic deal
– Deciding the direction of the next quarter
– Having a hard conversation with an unhappy client
– Choosing who to hire for a key role
Do not try to automate these. Do not apologise for them. They are your job. The point of automating everything else is to give you the bandwidth to do these well.
How to Score Which Tasks to Automate First
The four categories tell you what is possible. The scoring tells you what to do first.
For each task on your list, score it on three dimensions: frequency, time cost, and automation fit. Add the scores. Start from the top.
Frequency (1 to 5)
How often does this task happen?
– 1 = once a month or less
– 3 = once a week
– 5 = every day or multiple times a day
High-frequency tasks compound faster. A task you do every day that takes 10 minutes is 50 minutes a week, or ~43 hours a year. A task you do once a month that takes 2 hours is 24 hours a year. The daily task is a bigger win even though each instance is smaller.
Time Cost (1 to 5)
How long does each instance take?
– 1 = under 5 minutes
– 3 = 15-30 minutes
– 5 = over an hour
Note: time cost is not just the time you spend executing. It includes the context switch before and the tail of mental load after. A “5-minute” task that pulls you out of deep work and costs 20 minutes of refocus is really a 25-minute task. Score it honestly.
Automation Fit (1 to 5)
How cleanly does this fit into one of the first three categories?
– 5 = Fully automatable (Category 1)
– 4 = Heavily assisted (Category 2)
– 3 = Supervised (Category 3)
– 1 = Human-only (Category 4, do not automate)
Your Automation Priority Score
Add the three scores. The highest numbers are where you start.
A daily task that takes 30 minutes and is fully automatable scores 5+3+5 = 13. Huge win. Do it first.
A monthly task that takes 2 hours and needs heavy assistance scores 1+5+4 = 10. Worth doing, but after the 13s are handled.
A weekly task that takes 15 minutes and is supervised scores 3+3+3 = 9. Worth doing, but lower priority.
A monthly strategy decision that takes 3 hours and is human-only scores 1+5+1 = 7. Do not automate. Just protect the time.
This sounds mechanical because it is. The point is to stop arguing with yourself about which task feels most important and start ranking objectively. Feelings are noisy. The math is not.
The Quick Wins You Almost Certainly Have
Most businesses I audit have the same three or four quick wins sitting untouched. If you are looking for where to start, these are where to look first.
Lead Response Within 90 Seconds
If you generate leads and do not respond within the first few minutes, you are losing money. The research is unambiguous: 78% of deals go to the first responder (Harvard Business Review has covered this extensively). Most businesses take 4+ hours to first contact. Some take days.
This is a Category 1 task. A system can contact every new lead within 90 seconds by SMS, email, or voice. It can qualify them with a short series of questions. It can book an appointment if they are qualified, or route them to a human if they are not. No judgement required on the first touch. The judgement comes later, in the conversation that follows.
If you have any meaningful lead volume, this is almost always your highest-scoring task. See lead response time and speed to lead for the full case.
Database Reactivation
You have a CRM with thousands of contacts. You have never properly worked it. There is probably $50k to $500k sitting in there in unconverted leads from the last 24 months. Nobody on your team has time to call through the list, and your automated “we miss you” emails have a 0.3% response rate.
This is a Category 3 task. An AI reactivation system sends conversational multi-touch sequences across SMS and email, responds to replies, qualifies interest, and books appointments with anyone who bites. You review the results weekly. James, the finance broker, recovered $49,000 from 319 dormant contacts his team had already written off. Same list. No new ads. No new leads. Just a system that actually worked the database.
See database reactivation, reactivate old leads, and win back old customers for how this plays out.
Inbound Call Handling
Your phone rings. Your receptionist is on another call, or it is after hours, or they are at lunch. Dr Claire (dental practice) measured this: 47% of calls were going unanswered despite two receptionists. That is not a staffing problem. That is a systems problem.
Category 3. An AI receptionist answers every call 24/7, handles common requests (bookings, cancellations, general questions), qualifies new enquiries, and sends a summary to the human team. Dr Claire went from 47% missed to zero missed and saw 44% more booked appointments. Justin Touyz saw a 27% revenue boost in his first month. Donna Loeffler doubled her sales in the month of deployment.
Meeting Notes and Action Items
You sit in a meeting. You take notes. You forgot to distribute them. Or you distribute them, and nobody acts on them. Two weeks later, you are in the same meeting again because nothing has progressed.
Category 2. Tools already record and transcribe meetings (Fathom, Fireflies, Otter). An AI layer reads the transcript, extracts action items, assigns them to the right people, and reports back on what did or did not move. You spend five minutes reviewing instead of 45 minutes writing and chasing.
What Most Founders Get Wrong About Which Tasks to Automate
Here is the pattern I see repeatedly when helping founders decide which tasks to automate. The mistakes are consistent.
Mistake 1: Automating the Hardest Thing First
The most annoying task is usually the most complex one, and the most complex one is usually the worst place to start. Complex tasks involve judgement, context, and edge cases. They take longer to set up, break more often, and erode confidence when they do.
Start with a Category 1 task that has been eating small amounts of time daily. Something boring. Something nobody would write a case study about. Lead response. Follow-up emails. A daily report. You want a clean win that takes five minutes to notice and holds up over weeks. That is the win that makes you trust the system with something harder.
Mistake 2: Trying to Automate Across the Whole Business at Once
The instinct is to map everything, then build everything. This is how three-month projects become twelve-month projects that never finish.
Automate one task. Get it working. Measure what it saved you. Then pick the next one. Build the muscle of shipping small automations rather than designing a perfect system.
Mistake 3: Confusing Automation with Replacement
Automation is not about removing humans. It is about removing the repetitive parts of human work so humans can focus on the parts that actually need them. A receptionist freed from answering basic booking calls can spend more time handling difficult patients in person. A salesperson freed from data entry can spend more time in actual sales conversations.
If you are thinking about automation as “how do I fire people,” you will build the wrong systems and end up with worse outcomes. Think about it as “how do I make the people I have 3x more effective?”
Mistake 4: Trying to Automate Category 4 Tasks
Every few weeks, someone asks me to help them automate a task that is clearly human-only. Hiring decisions. Strategic planning. Handling difficult client conversations. You can automate the admin around these (scheduling interviews, drafting meeting agendas, preparing client summaries), but the judgement at the centre stays with you.
When you try to automate judgment, you either end up with bad decisions or you end up with so many override rules that the system is slower than just doing it yourself. Protect Category 4. Automate everything around it instead.
Mistake 5: Ignoring the Tasks Your Team Handles
Your time is not the only time that matters. If your ops manager spends six hours a week on something that could be automated, that is six hours of capacity you are renting. Map tasks across the whole business, not just yours. The scoring works the same way. The wins are just as real.
From Scoring to Shipping: The First 30 Days
Here is what this looks like in practice. If you score your tasks on Monday, what should your week look like?
Week 1: Audit
Spend 90 minutes listing every recurring task across the business. Daily, weekly, monthly. Yours and the team’s. Be honest about how long each one takes and how often it happens. The typical count is 50-100 tasks. If you list fewer, you are missing things.
Apply the scoring. Rank the list from highest to lowest. Look at the top 10.
Week 2: Pick One
From the top 10, pick one. Not three, not five. One. Ideally, a Category 1 task that scores 12+. Your first lead response automation is a good bet. So is a daily metrics summary. So is a missed-call text-back.
Get it built, tested, and running. Measure what it took and what it saves. This is your proof-of-concept for yourself and your team.
Weeks 3 and 4: Build the Rhythm
Pick the next two tasks from the top 10. Build them. By the end of the month, you have three automations running, and you know what it feels like to ship these. You also have a clearer sense of which tasks are worth the effort and which ones you want to leave alone.
Track your task automation percentage. If you had 80 recurring tasks and you have automated 3, you are at ~4%. The first 20-30% is where most founders start to feel the difference. The 180-day target is 60-70%.
Why This Belongs Inside an AIOS
Here is where this connects to the bigger picture. Automating individual tasks is valuable, but automated tasks in isolation are a collection of tools, not a system. The difference is context.
A lead response automation that does not know your business will sound generic. It will qualify badly. It will send the wrong leads to the wrong people. But a lead response automation that sits inside an AI Operating System, where your business context, your pipeline, your offers, and your team structure are already captured, will behave like a team member who has been there for years.
This is the difference between “I use AI to automate tasks” and “my business has an AIOS that runs on AI.” The first is a productivity hack. The second is a fundamental change in how the business operates.
If you want the longer read on this, see ai operating system and ai operating system for business. If you are earlier in the process and wondering where to start, where to start with ai is a better entry point.
The shortest version: you install context first (the AI knows your business), then data (the AI sees your numbers), then intelligence (the AI briefs you daily), then automate. Trying to automate without the first three layers is why most automation projects plateau. The task runs, but the output is generic, and you end up doing half the work anyway. With the full stack behind it, the same automation produces output that feels like it came from someone who already knows the business.
The Scoreboard You Actually Want
Every decision about which tasks to automate should ladder up to three numbers.
Away-from-desk autonomy. For hours per day, you can step away and nothing breaks. This is the real measure of whether your automation is working. If you automate five tasks but still cannot take a Friday off, you are automating the wrong things.
Task automation percentage. The scoreboard. Start at 0%. Track monthly. Watch it climb. The first 20-30% is revelatory. The 60-70% target is where the business genuinely runs without you being the hub.
Revenue per employee. Total revenue divided by total team size (including contractors). This is the number that tells you whether the system is creating real leverage. A business with $1M in revenue and 2 people is structurally different from a business with $1M and 8 people. Not better or worse as a business, but very different as a machine. Automation is how lean businesses get leaner without cutting muscle.
Every task you score and automate should be a vote for one of those three numbers. If the task does not move any of them, either it is not worth automating, or you have picked the wrong metric.
The Audit Is the Hard Part
The scoring framework is simple. What is hard is doing the audit honestly.
Most founders try to list their tasks from memory and produce a list that is 40% of their actual task load. The list skews toward the tasks they remember hating, and undercounts the small, invisible admin that actually eats the most time. The five-minute checks, the re-sending of attachments, the “just one more thing” requests from the team, the context-switching.
If you want an accurate list, you need a system that watches what you actually do. That is what an AI operating system is for. The context layer captures who you are and how you operate. The data layer watches your actual calendar, emails, and outputs. The intelligence layer synthesises this into a picture of where your time is going. Then you do the task audit with real data instead of guessing.
Without that foundation, you are scoring tasks that might not even be the right tasks. With it, the scoring is trivial because the list is complete.
Where This Leads
The businesses that get the biggest returns from AI aren’t the ones with the most automations—they’re the ones that know exactly which tasks to automate first. By focusing on recurring, high-impact work instead of chasing the most complex problems, they build momentum, recover bandwidth, and create a system that compounds over time.
Treat automation as an ongoing discipline, not a one-off project. Regularly audit your recurring work, score each task objectively, and implement the highest-value opportunities one at a time. Each successful automation creates more capacity to improve the next, turning small operational wins into meaningful strategic advantages.
The goal isn’t to automate everything. It’s to remove the repetitive work that doesn’t require human judgement, so you and your team can spend more time on the decisions, relationships, and ideas that actually grow the business. That’s where automation delivers its greatest return.
Next Step
If you’d like to map this out for your specific business, book a 30-minute Discovery Call. I’ll walk you through what AI could realistically take off your plate, how to roll it out properly at your size, and whether there’s a fit. No pitch, no obligation.
If you are not ready for that yet, start by looking at the tasks you did yesterday. Count how many fit Category 1. That number, the ones you could cross off permanently this month, is usually bigger than you think.
Frequently Asked Questions
Which tasks should I automate first?
Start with recurring, rules-based tasks that happen frequently and require little or no human judgement. Lead follow-up, appointment confirmations, reporting, and missed-call responses are often among the highest-impact opportunities.
How do I decide which automation will deliver the biggest return?
Score each recurring task based on how often it occurs, how much time it consumes, and how suitable it is for automation. The tasks with the highest combined scores should be your first priorities.
What tasks should never be automated?
Strategic decisions, hiring key team members, difficult client conversations, pricing major deals, and other activities that depend heavily on judgement, relationships, or creativity should remain human-led.
Should I automate multiple processes at the same time?
No. Focus on one high-value automation first, confirm that it delivers measurable results, and then build momentum by moving to the next priority. Small wins compound much faster than large, complex projects.
Does automation replace employees?
The goal is not to replace people but to eliminate repetitive work, so your team can focus on higher-value activities. Well-designed automation improves productivity without reducing the importance of human expertise.
Why is business context important for automation?
Without business context, AI produces generic outputs that often require manual correction. When automation understands your products, customers, processes, and goals, it behaves more like an experienced team member and delivers far more useful results.
About Octavius
Titus Mulquiney is the founder of Octavius AI, where he builds AI brains and AI workforces for founder-led businesses stuck running everything out of their own head. Twenty years in marketing, ex-Sony product manager, ex-GM Zeal NZ. Based in Auckland, working with operators across NZ, Australia, and the US. Connect on LinkedIn.