Compound field notes / AI assessment
Before you build automation, check what you already own
Your existing software may need a proper setup. Compare buying, connecting tools, and custom automation before paying for a build.
5 min read
In this guide
Check the software you already pay for before commissioning a build. The feature you need may be sitting in a settings page, waiting for someone to connect it to the way your team works.
I sell automation builds. I still want the assessment to recommend an existing tool when that solves the problem. A custom system becomes something you have to own, review, and maintain. It has to remove enough work to earn that responsibility.
There's room between buying a product and building an entire system, too. If the tools do their jobs and someone is stuck carrying information between them, connecting those tools may be all the work you need.
The decision before the build
Check all five options
Configure what you own, buy a product, connect existing tools, build custom automation, or keep the task manual.
Follow one job from start to finish
Pull up a recent job. Look at where it arrived, who touched it, and where someone had to copy, check, or chase something before it was finished.
“We need AI for quotes” could mean several different problems. In an illustrative example, a person checks open quotes every Friday, works out which customers haven't replied, and prepares follow-up drafts. Now you've got a task you can inspect.
Check the quoting tool before adding anything:
- Does it already have a follow-up feature?
- Does it know which customers replied?
- Can a person review the drafts before they go out?
The answers decide how much work remains.
Try the existing feature on a few records you can safely use. Include one incomplete record and a customer who has already replied. A clean demo won't tell you whether the tool creates more checking for your team.
Pick the smallest change that handles the whole task
| Choice | When it fits | Check before committing |
|---|---|---|
| Configure a tool you own | The feature exists and fits the task | Setup, permissions, training, and how it handles your records |
| Buy a product | Your process is close to a workflow the product already supports | Subscription cost, adoption, and access to your data |
| Connect existing tools | Each tool works, but the handoff is manual | Customer matching, duplicates, and failed transfers |
| Build custom automation | Important steps remain that existing options don't handle well | Expected return, review work, and ownership after launch |
| Keep it manual | The task is rare, changing, or inexpensive to complete correctly | A clear owner and a reliable way to check it's done |
You can use several of these in the same business. Buy scheduling software, connect bookings to your customer records, and leave unusual customer requests with a person.
A request that needs the owner's judgment can stay with the owner. The routine work around it can still move faster.
A connection still needs someone to handle failures
Say one tool records a new customer and another needs that information to start a job. A connection can carry the record across. It also needs:
- Customer matching: a rule for finding the right customer in both tools.
- Duplicate checks: a way to avoid creating the same job twice.
- Visible failures: somewhere your team can see that a transfer didn't work.
Someone should be able to answer, “What happened to this job?” without checking four different systems.
Measure the whole task. Saving ten minutes of copying doesn't help if the team spends twenty minutes figuring out whether the records arrived correctly. That's an illustrative comparison, but it's a useful test when you're watching a demo.
The Newgarden case study shows a focused check-photo-to-invoice workflow that recovered roughly six hours a week. The value came from dealing with a particular piece of recurring work. Your own records, volume, and review time determine what's possible in your business.
Custom work needs a financial case and an owner
Look more closely at a custom build when the work repeats, the inputs are clear enough to use, and you can recognize a correct result. Then work out what solving it would be worth.
Compound scopes and prices builds around a target of 5–10x return within the first year. That means expected first-year benefit of five to ten times the total first-year investment. We test the assumptions in the assessment and agree on a fixed scope and fee. The target isn't a guarantee.
Include software, usage, upkeep, and the checking your team will still do. The pricing guide explains how we calculate the multiple and separate recovered capacity from cash savings.
Name the person who owns the workflow after launch. They need to know how to pause it, where exceptions show up, and how to get help. Settle that before building.
When comparing proposals, ask the vendor to:
- Show a failure: walk through an incomplete or failed task and what your team sees.
- Explain the next step: who fixes it, and is that work included in the price?
- Show the handoff: how would you export your records and give the system to someone else later?
Waiting can be the right decision
Hold off on a build if:
- The process changes every week.
- Your team can't agree on what a correct result looks like.
- Nobody has time to review the output.
Write down a few completed examples and settle the rules. Then test again.
That preparation gives an existing product a fair test and gives a builder something concrete to price.
The $2,500 AI Assessment takes 14 days and produces the ranked plan, cost model, tool choices, and first-build scope. The deliverables guide shows what you should be able to do with that plan.
If I don't find a durable, reliable solution to your bottleneck, you get the full assessment fee back. If a Compound build starts within 30 days of assessment delivery, the full $2,500 comes off its price. After that, the credit expires.
For the free discovery call, bring one task and the tools you're using now. We can start by checking what you already have.