Chad Rubin
May 10, 2026 · Updated May 11, 2026 · 12 min read
Operator notes by email
Short, opinionated takes on AI agents, Amazon PPC, pricing, and inventory. No fluff. About once a week.
I have been on both sides of this. I have hired agencies, fired agencies, run PPC in-house, and built software that competes with the agencies I used to write checks to. When a brand owner asks me whether they should fire theirs, I do not give the easy answer. There is no easy answer.
Most agencies still do real work. A handful are coasting on auto-bidding tools and a slick PDF deck. The trick is figuring out which one you have, and then deciding whether the work is worth what they charge in a year where AI handles more of the tactical layer than it did twelve months ago.
This is not a hit piece. I have referred friends to agencies this year and I will probably do it again. What I want to give you is the same checklist I would run on my own account if I were paying somebody outside my four walls to manage seven figures of ad spend.
Read this if you are paying anywhere from 5 to 15 percent of ad spend, or a flat retainer that has crept north of 5 figures a month, and you cannot tell anyone in plain English what you are getting back.
Some of it is real work that is hard to replicate. Some of it is autopilot work that decent software does for a tenth of the cost.
The real work, the part agencies still earn their fee on:
From reading to action
If the framework above sounds familiar, your Amazon account is probably carrying the same drag. Apply and we will show what Marko, Oracle, and Bruno would change in your first week.
Ran a 7-figure Amazon brand for a decade. Founded Skubana (acquired). Co-founded Prosper Show. 15+ years on Amazon.
Join the brands that replaced agencies and tools with AI employees.
The autopilot work, quietly automated for years:
If 80 percent of what your agency is doing lives on the second list, you do not need an agency. You need software and somebody on your team who can read the output. If the work is genuinely on the first list, you have a different decision to make.
"Since onboarding, our POAS and ROAS have both improved, driven by a much more disciplined, data-led approach to bidding and budget allocation."
Here is the decision framework I run when I am evaluating whether an outside team is still worth the spend. Five fork points. Each one has a clean answer. At the end you will land in one of three buckets: keep, replace, or augment.
Run all five. Do not stop at the first red flag and fire on impulse. The point is to see the whole picture.
This is the one almost everyone gets wrong, including me when I was hiring agencies.
PPC, pricing, and inventory are the same problem. If your price moves up 6 percent because your supplier raised costs, conversion rate drops, ACoS climbs, and campaigns start burning budget on impressions that will not convert. If inventory drops below 4 weeks of cover, you should be pulling back on top-of-funnel spend. If you launch a 15 percent off promotion, you should be leaning in harder.
Most agencies do not touch pricing. They do not touch inventory. They will tell you they "coordinate" but what they mean is they got a Slack message from your ops team three days after the price change.
Ask: when was the last time they made a bid adjustment because of a price change or inventory event, not because of an ACoS movement? If they cannot give you a specific example from the last 30 days, PPC is siloed. That is the single biggest argument for moving to a system where pricing, PPC, and inventory share the same operating brain.
When you signed the contract, you probably met a senior person. Maybe a founder. They were sharp. You felt good.
Six months in, who is actually touching your account? If you do not know, that is your answer.
Senior PPC operators are expensive. Agencies cannot put one on every account unless your spend is genuinely large. What happens in many shops is a junior gets handed your account along with twelve others. They run a weekly checklist. The senior person shows up on the QBR call to make it feel premium.
Ask for the name of the person doing the actual work this week. Ask how many other accounts they manage. If the answer is "12" and your spend is "average for the agency," your account is not getting senior-operator attention. It is getting checklist attention, and checklists can be automated for a fraction of what you are paying.
Ask your agency for a rough estimate of monthly hours on your account. Most will not give you a clean number because they are uncomfortable with the math.
Run the math yourself. A 7,000 dollar retainer at 8 hours a month is 875 dollars an hour. There are very few PPC operators worth that.
Hour count alone is not the disqualifier. What matters is hours-to-outcomes. If 6 of those 8 hours are strategic work you could not replicate (launches, competitive moves, rebuilds) you may be getting your money's worth. If 6 of those 8 hours are clicking buttons an automation could click, you are not.
Pull your last three monthly reports. Read them like you have never seen the account before.
Bad reports: "ACoS was 24 percent, down 2 percent. Sponsored Products spend was up 8 percent. Top performer was X." That is a description. They are reading you the dashboard.
Good reports: "We pulled spend off ASINs 7, 11, and 14 because conversion dropped after a competitor cut price. We moved 18 percent of that budget into ASINs 3 and 9 where you have a 4-week inventory buffer and margin advantage. Recommend we stay on this stance until competitor pricing normalizes."
The good version tells you what they decided, why, and what they recommend next. Short. Specific. Not trying to look impressive. If your reports are mostly the first version, you are paying for theater.
Three common fee structures, each with a hidden incentive.
Percentage of ad spend (commonly 5 to 15 percent). The agency makes more when you spend more. Fine when you are growing. It quietly punishes the agency for pulling spend back to protect margin. They will not say this out loud. The model is what it is.
Flat monthly retainer. Predictable for both sides. The hidden incentive is autopilot, because every hour they put in lowers their effective rate.
Performance-based (tied to TACoS, ACoS, or revenue). Sounds great, almost never works cleanly. The tied metric is usually ACoS or revenue, neither of which captures profit, inventory health, or category position. An agency hitting an ACoS target while your inventory rots is technically winning.
None of these are evil. They are incentive structures. Match the incentive to the goal you actually have. If your goal is profit not revenue, every fee structure on this list misses, and you should be running PPC closer to in-house.
Keep your agency when:
If you are keeping them, use this framework as leverage in your next renewal conversation. Ask for:
You will be surprised how many agencies will agree to all four. The good ones want this conversation. They are tired of being grouped with the autopilot shops.
Replace your agency when:
The mistake people make is replacing the agency with software alone. Software does not have a brain that decides what to launch, what category position to chase, or whether to push share or protect margin. Software executes. An operator decides.
The right replacement for most brands doing 1M to 50M is software plus one internal operations lead. The lead does not have to be a senior PPC manager. They need to be a generalist operator who can read AI recommendations, override what looks wrong, and translate quarterly business goals into automation rules. Dramatically cheaper than a full management agency, and far more aligned with your business because they are inside it.
For the longer treatment, see the no-employee Amazon business and AI operating system posts.
"We've been working with Profasee for over a year, using both their Repricer and PPC services. Their team is responsive, flexible, and thoughtful in the way they support each account."
This is the most underrated path and almost nobody talks about it.
Augment when:
Software handles bid adjustments, negative keyword harvesting, budget reallocation, dayparting, and the rest of the autopilot list. Your agency keeps the strategic seat: launches, rebuilds, Sponsored Brands and DSP, competitive intelligence, and the QBR conversation.
Renegotiate the fee. Either a 30 to 50 percent discount on the old retainer, or off percentage-of-spend entirely onto a smaller strategic retainer plus a project fee for launches.
If they are honest, they will not fight you. The good ones know the autopilot work is going to software either way. They want the strategic seat. The bad ones fight because the autopilot fee was keeping the lights on.
Whether you are replacing or augmenting, do not cancel on Friday and switch on Monday. Run in parallel for at least one full inventory cycle, ideally 60 to 90 days.
The trust ladder post walks through how to hand tasks off to AI agents safely. Do not skip the trust steps. Brands that go from zero to fully automated in a week are the same brands coming back six weeks later wanting their agency back.
The autopilot work on the second list at the top of this post is what Marko, our PPC manager AI employee, handles inside Profasee Ultra. Bid adjustments, negative keyword harvesting, budget reallocation, dayparting, structure cleanup, search term mining. The difference is coordination. Marko is not running PPC in a silo. He reads from the same memory as Oracle (pricing) and Bruno (demand planning).
That coordination is what fork 1 is about. When Oracle moves price up to protect margin, Marko sees it and pulls back top-of-funnel spend on the affected ASINs. When Bruno flags an ASIN heading into stockout in 4 weeks, Marko reduces aggression. Most agencies cannot deliver this because they do not have a seat at the pricing or inventory table.
Mission Control gives your operations lead the surface to run it. You see what Marko is recommending. Approve, override, or step back. Every decision logged. The trust ladder is real and visible.
See the PPC manager AI employee, Amazon PPC software, and pricing pages. If you want to see whether your catalog is a fit, apply here and we will tell you honestly. We have turned brands away because they were better off staying with their agency another quarter.
Run the five fork points: coordination with pricing and inventory, presence of a real senior operator on your account, monthly hours, report quality, and fee model alignment. If three or more fail, replace or augment. If most pass, renegotiate the contract using the framework as leverage. Do not fire on impulse. Run any transition in parallel for 60 to 90 days.
Software alone is almost always cheaper than a full management agency, often by 60 to 80 percent. But software alone is not a fair comparison. The right comparison is software plus one internal operations lead versus the agency. For most brands doing 1M to 50M, that combined cost is still meaningfully less than what they are paying an agency, with better coordination across pricing and inventory.
The real work is launch sequencing, competitive intelligence, brand strategy on Sponsored Brands and DSP, account-level escalations with Amazon, and translating business goals into campaign architecture. Bid adjustments, negative keyword harvesting, and budget reallocation are autopilot work that does not need an agency in 2026. If 80 percent of your invoice is autopilot work, you are overpaying.
Hire one operations lead (a generalist operator, not a senior PPC manager), give them software that handles the tactical layer, and run a 60 to 90 day parallel transition with your existing agency. Roll software forward category by category. Do not flip the whole account at once. The operator's job is to set the goals, read recommendations, override what looks wrong, and translate quarterly business strategy into automation rules.
AI can replace the tactical layer of a PPC manager, which is most of the day-to-day work. AI cannot replace the operator who decides what to launch next, what category position to chase, or how to weigh margin against share. The honest answer in 2026 is that the role is shrinking, not disappearing. One operator with strong AI tooling can do what a small team did three years ago.
Yes. Software without a human operator is the most common failure mode. Software executes, the operator decides. The operator does not need to be a senior PPC specialist. They need to be a generalist who can read the AI's output, spot when something looks wrong, and connect quarterly business goals to the rules the software runs by.
Common ranges are 5 to 15 percent of ad spend or a flat monthly retainer. The fairness question is not really about the dollar amount. It is about what work the fee actually covers. A 15 percent fee can be a steal if the agency is doing genuine strategic work and coordinating across pricing and inventory. A 5 percent fee can be a ripoff if it is paying for autopilot work that software does for a fraction of the cost. Match the fee to the work, not to a benchmark.
