On this week’s episode of the Area of interest Pursuits podcast, Corey Ganim and I focus on how to consider AI automation in a manner that really results in ROI. This subject has a selected give attention to AI brokers, reusable “expertise”, and why instruments like Claude Cowork are altering what automation appears like each day.
Corey runs Return My Time, and he comes at this subject from a builder’s perspective, not a buzzword perspective. He additionally is aware of our viewers is already previous the “what’s ChatGPT?” stage, so we spent the majority of the dialog within the weeds on frameworks, execution, and sensible examples.
Watch the Full Episode
Why Most AI Experiments By no means Flip Into ROI
Corey began by calling out one thing most of us have felt: there are too many instruments, an excessive amount of noise, and too many “cool demos” that by no means translate into higher enterprise outcomes. His repair is to cease judging AI by how spectacular it appears and begin judging it by what it improves in your online business.
That leads into his “three levers of ROI” framework, which is an easy filter you may apply earlier than you spend a minute setting something up. If a device, agent, or workflow doesn’t pull no less than considered one of these levers, Corey’s view is that it’s not price your time.
- Effectiveness: Does this aid you earn more money?
- Effectivity: Does this save significant time?
- High quality: Does this enhance the product or buyer expertise?
A Concrete Instance of “Effectiveness” ROI
Plenty of enterprise house owners begin with effectivity, as a result of saving time is the obvious win. Corey agrees that fundamental chatbot use might help, however he thinks the larger upside reveals up when AI helps you seize income you’d in any other case miss.
He shared an instance of an AI “velocity to steer” agent that responds to inbound leads rapidly with a personalised response. Within the interview, he referenced a stat that you’re “21 occasions extra doubtless” to land a buyer should you reply inside 60 seconds versus an hour or two. That is precisely the type of leverage that may influence income, not simply workload.
- Native service companies are particularly delicate to response time as a result of buyers typically contact a number of suppliers.
- “Velocity to steer” workflows could be applied with out changing your full gross sales course of; they only take away the delay.
- In case you already purchase leads or run paid adverts, a quicker response improves the economics of what you might be already spending.
The place to Begin: Audit, Optimize, Automate
After you resolve what outcomes matter, the subsequent downside is determining what to automate first. That is the place Corey launched his AOA framework: audit, optimize, automate, in that order.
His level was that most individuals bounce straight to automation as a result of they’re annoyed with a process, however that may lock in a bloated course of. In case you automate a 12-step workflow that might be diminished to six steps, you haven’t actually improved the system; you will have simply buried inefficiency beneath automation.
- Audit: Establish what you might be truly spending time on (calendar + to-do record is an effective beginning supply of reality).
- Optimize: Scale back steps, take away waste, simplify handoffs, make clear what “accomplished” appears like.
- Automate: Solely after the method is cleaner do you have to hand it to an agent or automation device.
Utilizing AI to Optimize Earlier than You Automate
One of many extra helpful elements of this dialog was the concept AI might help with the optimization step, not simply the automation step. Many people assume we have to be the professional who redesigns a workflow, however Corey’s view is that AI is usually extra goal than we’re as a result of it has no attachment to “how we’ve all the time accomplished it”.
He described a sensible option to do it: mind dump your present course of in plain language, let the AI suggest a slimmer model, after which take that optimized model ahead into automation. Corey used his personal weekly Costco grocery ordering for instance, the place optimization lower out steps earlier than he handed the workflow to an agent.
- A small discount in steps can matter greater than it sounds, as a result of recurring duties compound over weeks.
- Voice-to-text could make “course of seize” quicker than typing if you end up documenting a workflow.
- Optimization can be the place you make clear constraints, like what inputs are required and what the output format ought to be.
Claude Cowork: What It Is and Why It Feels Completely different
Claude Cowork represents a shift from “AI solutions questions” to “AI does work”. Corey framed it in a memorable manner: a standard chatbot is reactive, whereas an agent is extra like a chatbot with palms that may work together along with your instruments.
He defined that Cowork can hook up with instruments like electronic mail, Google Drive, CRMs, and process managers, then carry out actions inside them. The interface nonetheless appears like a chat, however the capabilities change as a result of it might execute actual duties and run scheduled workflows.
- Software connections are the place lots of worth seems, as a result of context now not lives solely contained in the chat window.
- Scheduling issues as a result of it turns AI from “one thing you keep in mind to make use of” into “one thing that reveals up on time”.
- Execution plans are helpful as a result of they pressure construction earlier than an agent begins clicking round in your methods.
A Sensible Cowork Use Case: Higher Gross sales Calls With Much less Prep
One instance Corey gave was for a buddy working an actual property teaching enterprise with a high-ticket provide (he talked about a $13K ticket dimension). That buddy was taking 5-7 gross sales calls per day and spending 10-Quarter-hour earlier than every name scrolling by way of CRM historical past to recollect context.
Corey’s advice was to attach Cowork to the CRM and have it mechanically ship a abstract a couple of minutes earlier than every name, together with context from prior conversations and dialog starters. The end result is that as a substitute of getting ready 10-Quarter-hour forward, you may put together in seconds with a look at a briefing message.
- This type of workflow scales properly to a staff as a result of it standardizes preparation throughout a number of reps.
- The advance will not be solely time financial savings, however it might additionally cut back “chilly begin” awkwardness firstly of calls.
- Summaries can embody key constraints like price range, timeline, and objections that you really want top-of-mind.
Brokers vs Abilities: The Lacking Hyperlink for Individuals Who Really feel Caught
An enormous theme of the episode was why many savvy AI customers nonetheless really feel like they don’t seem to be getting the promised time financial savings. You could be good at prompting and nonetheless get trapped in a loop: paste an electronic mail, rewrite it, shorten it, change tone, paste once more, and ultimately it looks like you could possibly have written it your self.
Corey solutions that expertise are what make AI reusable and constant. In his framing, the agent is the doer, whereas the ability is the recipe the agent follows for a selected process, step-by-step, in the identical manner each time.
- Abilities cut back “immediate drift” as a result of you aren’t reinventing the directions on each run.
- Abilities could be shared throughout instruments as a result of the idea is spreading throughout the AI ecosystem.
- The ability strategy is very helpful for any work you do weekly or every day, as a result of repetition is the place the ROI reveals up.
The Foundational Talent Corey Recommends Everybody Construct First
If Corey may give just one subsequent step, it was this: construct your model voice ability. He argued that it’s the basis beneath many different expertise, as a result of it makes outputs persistently sound such as you with out repeated back-and-forth modifying.
He additionally stated it is not sophisticated to arrange, and that he constructed his in about 20 minutes by giving examples of his writing, plus transcripts that present how he speaks. From there, he described utilizing Claude’s skill-building functionality to create and set up the ability, so it turns into the default “voice layer” throughout future work.
- Inputs that assist a model voice ability: writing samples, electronic mail examples, podcast transcripts, and a brief record of “phrases I keep away from”.
- A model voice ability will not be restricted to advertising. It additionally improves inner communication, like staff updates and SOP drafts.
- After getting this ability, you may pair it with different expertise so your entire system stays constant.
Excessive-Impression Abilities Most Enterprise House owners Can Use Instantly
After we dug into the idea, I requested Corey for examples of expertise which can be extensively relevant. He talked about model voice first, then pointed to widespread locations the place enterprise house owners spend time and a spotlight: electronic mail, every day planning, and sales-call context.
What stood out to me is that none of those are “flashy automations”. They’re the type of recurring, barely annoying duties that quietly eat hours every week, which makes them good targets for skill-based automation.
- Electronic mail drafting ability: A default format and tone so routine messages come out usable instantly.
- Each day transient ability: A scheduled morning abstract of calendar gadgets and key messages to deal with.
- Gross sales name prep ability: A pre-call “battle card” pulled from CRM historical past and exterior context like a LinkedIn abstract.
High quality Assurance: Easy methods to Belief Outputs With out Babysitting Them
One pushback individuals have is, “How do I make certain this doesn’t produce rubbish?” Corey’s strategy was to deal with expertise like software program: model them, take a look at them, and refine till you belief them.
He stated V1 isn’t good, however it’s typically 80% of the best way there, after which V2, V3, or V4 is the place it turns into reliable. The bottom line is to bake QA into the ability creation course of by working it, reviewing the output, and adjusting the directions till the output format and high quality match what you want.
- Examples of QA tweaks embody altering output format (Phrase doc vs Excel), tightening construction, or including required sections.
- Testing is quickest if you use practical inputs, not idealized examples.
- As soon as a ability is secure, you usually solely replace it when necessities change.
What “At Scale” Can Truly Look Like in a Week
Corey shared a private metric that helps make this actual. He described utilizing Todoist to determine recurring duties, then difficult himself to show these duties into expertise so he wouldn’t should do them manually once more.
After about two and a half months of doing this, he stated he had outfitted his AI agent with round 42 expertise. The essential half was not simply the quantity, however the concept every new ability represents a process he can now offload absolutely, relatively than partially.
- It is a compounding system, as every new ability reduces future workload and will increase future capability.
- Recurring duties are the candy spot as a result of they repay the setup time again and again.
- A “expertise stock” can develop into a playbook you may hand to future hires or contractors.
Content material Automation That Nonetheless Performs: Corey’s X Article Instance
Corey additionally gave a efficiency instance that caught my consideration, as a result of it connects automation to actual outcomes. He described constructing an “X article author” ability that generates an 800-1200-word article in his voice, plus a number of headline and thumbnail choices, turning a two-hour course of into about 10-Quarter-hour.
Much more fascinating, he stated an article he posted the day earlier than the interview was absolutely AI-generated with no edits and reached 3.6 million views about 36 hours after posting. That could be a robust reminder that “automated” does not mechanically imply “low high quality”, particularly when expertise and voice constraints are dealt with properly.
- If a platform is rewarding a format, like long-form articles, automation might help you publish persistently sufficient to profit.
- A number of headline variations matter as a result of titles are sometimes the very best leverage variable in content material efficiency.
- One of the best use instances often mix analysis, construction, and model voice, relatively than counting on uncooked era alone.
Corey’s Provide and an ROI Snapshot
Corey briefly outlined what ReturnMyTime.com does immediately, together with an “AI instruments evaluation” the place he interviews a enterprise proprietor, identifies bottlenecks, and prescribes off-the-shelf instruments that may save time. He described a goal of shopping for again 5-10 hours per week, with a refund assure if they can not discover no less than 5 hours per week in financial savings.
He additionally shared a median value snapshot: about $42 monthly in instruments paired with roughly six hours saved per week, which he framed as a easy “what’s your time price?” calculation. Whether or not somebody buys assist or does it solo, the takeaway is that small month-to-month device prices could be trivial in comparison with the worth of reclaimed hours.
- He pointed listeners to a brief “AI literacy” quiz at returnmytime.com/quiz that recommends sources primarily based in your degree.
- The evaluation mannequin is fascinating as a result of it prioritizes off-the-shelf adoption, not customized engineering from day one.
- Time-saving estimates develop into extra plausible when they’re tied to particular workflows, not imprecise “AI will assist”.
Remaining Ideas
The large win from this episode is the shift from “utilizing AI” to constructing an AI working system for your online business. Corey’s frameworks maintain you targeted on outcomes first, then information you thru a wise construct order so you don’t automate chaos.
In case you’re overwhelmed, the only transfer is the one Corey saved repeating: construct the model voice ability first, then begin changing your recurring duties into expertise one after the other. Whenever you strategy AI this manner, you aren’t chasing instruments; you might be steadily eradicating work out of your plate and changing it with reusable methods.
