Phill

Philomath
Problem Solver
Strategist
Thinker

Tech & Engineering Thoughts and Musings
Founder @TheBluebearAI &

[Phill @philjoeco]

@SinaShayesteh I'd keep the reason each deal stalled with the lead. A result from a similar company is useful when it answers that concern. If they paused because implementation looked risky, another revenue screenshot probably won't change their mind.

[Phill @philjoeco]

@raju_GoogleAds What takes most of that review: checking the account data or deciding whether the recommendation makes sense for the client? Linking each finding to the figures it used could reduce the first part while keeping the judgment with you.

[Phill @philjoeco]

@chrisreedbates History can tell the agent what a good target looks like. It can't prove that any of the 27 qualifies. I'd ask for the source and evidence against each criterion, with gaps left unknown. Then the review is about the targets, rather than whether the agent did the work.

[Phill @philjoeco]

@thedanielbudai The tracking fix deserves a note in the client report: what counted as a conversion before, what counts now, and when it changed. Otherwise removing non-sales conversions can make the next report look worse even though the measurement is finally useful.

[Phill @philjoeco]

@iGrowEcomBrands Being in the room before decisions are made seems hard to delegate. How do you pass that understanding to your team? Keeping the reason behind a client decision alongside the brief could help them make the next call without everything coming back to you.

[Phill @philjoeco]

@mfishbein How often do you check the clips it throws away? I'd include some rejected suggestions in the review sample. Otherwise you can see what it chose badly without noticing the good clips it stopped showing you.

[Phill @philjoeco]

@itsjacobbreton What still lands with Will each week after onboarding: routine status checks or client-specific decisions? I'd separate those before automating the next step, so the repeatable work goes away without sending every unusual case back to him.

[Phill @philjoeco]

@Brands_vant For video, I'd define a revision round as one consolidated set of feedback from an agreed approver. Otherwise three people commenting on different cuts can use up the time without anyone agreeing which version is final.

[Phill @philjoeco]

@eugene_ego Your point about the unread notification is where I'd make the rule concrete: who must acknowledge it, by when, and who takes over if they don't. Otherwise 'escalated' can become a status that hides work nobody has accepted.

[Phill @philjoeco]

@itsheatherhop The changes you make after reading it aloud seem like the most useful feedback for AI: which line sounded unnatural, what you replaced it with, and why. Saving just the final script loses the judgment that got it there.

[Phill @philjoeco]

@chriswiser The useful test is whether a real job has moved out of the old process. People can log into a new tool every day and still do the work in spreadsheets. Before a longer commitment, I'd want to see the team complete one recurring job without doing it twice.

[Phill @philjoeco]

@SKNymuzzaman I'd track how long each script waits for approval separately from time spent writing it. If waiting is the bigger delay, faster production just builds a larger queue. Which is harder in your workflow: getting a decision or working out which feedback is final?

[Phill @philjoeco]

@redstonefndry I'd add a handoff test: can the incoming owner receive an alert and restore a failed form without asking the builder? The account list tells you who has access. That exercise shows whether the next person can actually run it.

[Phill @philjoeco]

@connor_hacks Thanks, that clarifies who makes the fixes. I'd keep each finding linked to the MSP's evidence of remediation, so the handoff for retesting shows what's ready to check and what's still open. That saves both teams reconstructing it from email.

[Phill @philjoeco]

@tryapexweb I'd give a range first, based on the booking rules the client describes, then confirm a fixed setup price after checking them. If that check needs custom integration work, I'd scope it as paid discovery rather than do the build just to price it.

[Phill @philjoeco]

@directorcia That helps. I'd track why each proposal comes back for rework too: missing client inputs, a changed scope or an error in the draft. That would tell us whether AI is saving work or just moving it into review.

[Phill @philjoeco]

While reviewing Pix2Code's sales feedback, we found won deal value being treated as revenue.

For an agency, winning the work, getting paid and making money on delivery are different things.

An AI choosing which leads to chase needs that distinction too.

[Phill @philjoeco]

Our demo narration passed the missing-word check: 15 out of 15.

That still didn't tell me whether I'd want a client to hear it.

Working on Pix2Code keeps exposing this gap: checking that AI content is complete isn't the same as deciding it's ready to send.

[Phill @philjoeco]

3/3
But someone still needs to understand the business, decide what to build, check the work and own what breaks.

I think that changes the opportunity for engineers and agencies: an ongoing technical relationship, with someone accountable for the result.

[Phill @philjoeco]

1/3
I think agentic software engineering is doing more than shrinking teams. It’s collapsing several roles into one.

For a small company, one engineer could become its technical lead, combining parts of the developer, forward-deployed engineer and CTO roles.

[Phill @philjoeco]

2/3
That person could run a team of agents to build and maintain the company’s software.

With enough leverage, they could become a fractional CTO across several businesses—giving smaller companies access to technical leadership they couldn’t previously afford.