An MSP owner put a question to Josh Peterson recently that has an obvious answer right up until you look at it closely. The company spends somewhere between fifty and sixty thousand dollars a year on an appointment setter, a capable one, better managed than most. An AI outreach service had quoted them thirty-six thousand for the same job, and by the vendor's account it would do more of it, more intelligently, across far more of the market. The owner takes the sales call either way, so the capacity on the back end does not change. On price alone the decision writes itself, which is exactly the problem. Price is the only part of this decision that is legible before you buy. Everything that determines whether it works sits behind vendor language about scrubbing, sequencing and ICP matching, and an owner who has never bought AI marketing services has no way to see through it. That is not a gap that reading vendor websites will close. It is a question about what you are actually purchasing, and the answer has less to do with the technology than with whether your ideal client profile and your marketing strategy can survive contact with a channel that fires ten times faster than the one you have now.
The technology question turns out to be the easy one, and it is settled. The models crossed a line in the last several months, and a competent AI system can now assemble a better list, write a better message and follow up more intelligently than a small team doing it by hand. What has not changed is that a vendor controls only the part of the process that ends when an email lands in an inbox. Everything past that point belongs to the company that hired them. So the entire evaluation collapses into one question the buyer has to ask before the first demo, and most never do: what outcome am I paying for, and can this vendor show me how their work reaches it? An owner who names an activity will be sold an activity, at which point the vendor has delivered exactly what was requested and the pipeline is still empty. The leverage in this decision is not in the negotiation. It sits in the specificity of the goal you bring to it, and in whether the message you are about to amplify has ever been tested. That is the same discipline that separates managing sales activity from managing a number, and it is why the cheaper line item is not automatically the better decision.
The numbers in front of that owner were fifty to sixty thousand against thirty-six. One number buys a person who books appointments. The other buys a system that promises to book more of them across a wider slice of the market. The owner takes the meeting in either case, so the sales capacity behind the appointment is identical and the comparison looks clean. It is not clean. An appointment setter is a known quantity with observable output and a manager who can coach it, and when it underperforms you can listen to a call and hear why. The AI service is a system whose observable output, by the vendor's own reporting, is opens and clicks. Those are not the same kind of number. Putting them beside each other on a spreadsheet implies a comparison that has not actually been made.
This is where most of these decisions are really decided, well before anyone evaluates the technology. A cheaper line item with a worse feedback loop is not obviously the better purchase, and the P and L will not tell you which one you bought for at least two quarters. The only way to make the comparison honest is to insist that both options are measured against the same thing, which means picking that thing first.
Gary Boyle's opening position is blunt: most marketing companies are liars. Then he takes his own accusation apart, and the second half is more useful than the first. A marketing vendor can influence who receives a message, what it says, when it arrives, and what happens on the second and third touch. It cannot influence anything that happens after the message lands, which is precisely where the decision to book a meeting gets made. So when a vendor defines a click as a lead and reports that click as delivery of the service, they are not necessarily lying. They are reporting the boundary of what they control. The dishonesty enters only when nobody in the room says out loud that the boundary is not the goal.
That reframing makes the evaluation practical rather than technical, which matters because the technical question is not one a first-time buyer can answer. You will not learn to assess list-scrubbing methodology in a procurement cycle. What you can do is state the outcome first and make the vendor answer against it. Gary's version is close to a template: my goal is ten first-time appointments a month, here is the ideal client profile, here are the qualifications, what can you do to help me reach that? A vendor who traces their work toward that number, or who tells you honestly which part of the gap they cannot close, is a vendor you can manage. A vendor who redirects to open rates has answered a different question, and the fact that you can hear the substitution is the whole value of asking.
Josh Peterson's concession partway through the conversation is the part worth sitting with. He rarely wants to call vendors liars, he says, and then he points out that the buyer lies to himself at least as often. The vendor says a click is a signal. The buyer, who wants the problem solved and wants it solved for thirty-six thousand dollars, hears that and says it sounds right. Both parties leave the meeting having agreed on something neither of them believes will produce clients. Nobody deceived anybody. The optimism was mutual, and it was load-bearing.
This is why the fix is a discipline rather than a better vendor. Walk in saying you need to send more emails and a company that sends more emails is a perfect fit; the engagement will be judged a success while nothing changes in the pipeline. Walk in saying you need ten first-time appointments a month from companies that look like this, and the same vendor has to either show you the path or decline the work. Gary's formulation is exact. If you do not start with the outcome you are looking for, they actually cannot lie to you, because the lie needs a vague goal to attach itself to. The vagueness is supplied by the buyer, every time.
Gary has spent roughly thirty hours of his own time, and by his estimate around ten times that in machine time, reworking how Bering McKinley talks about Vision. That is a strange place to put that much effort when the tooling is the interesting part. He put it there because of what surfaced when he started running the company's own calls through analysis. His message, Josh's message and the appointment setter's message are not the same message. They say different things at different times, and a share of the difficulty in those calls is self-inflicted rather than handed to them by the prospect.
That finding generalizes uncomfortably. For twenty years the binding constraint on outbound was capacity, so the rational response to a thin pipeline was more activity, and messaging was something a company settled once and stopped examining. AI removes the capacity constraint. What is left is the message, delivered now at ten times the volume to a recipient whose own filters are improving at the same rate. An MSP that outsources both the message and the channel and then hopes has not bought leverage. It has bought amplification, which is only worth having if the thing being amplified is correct. The honest sequence is the message first, then the channels, then a way to assess yourself against both, then iteration. Buying the channel first is how a company spends fifty thousand dollars proving that its positioning was never tested, and the proof arrives two quarters late.
In the last stretch of the conversation Josh asks Gary to explain the rules behind the workflows that flag his aging deals, and the answer is not really about the CRM. Prompting an AI to help you do better at something does not work, Gary says, because it will tell you a different thing every time you ask. What works is a framework it can enforce. Bering McKinley's is one sentence: no deal sits in the deciding stage longer than thirty days. Past that, the deal is dead and gets closed out. Josh resists it in real time, which is the useful part of the exchange, because his resistance is the normal one and he is honest about where it comes from.
The rule is doing two jobs at once. It clears the pipeline of deals that exist mainly to make the pipeline look healthier than it is, and it forces a change in message at day thirty rather than a fourth version of the same follow-up. Killing the deal is not the same as abandoning the relationship. The contact stays, enters a nurture sequence, gets a six-month touch and keeps an owner. What dies is the fiction that it will close. The broader point survives the specifics of anyone's CRM: every workflow worth automating sits on top of a business rule that a person had to decide, and the reason most automation produces noise is that the rule was never written down. AI did not remove the need for that decision. It made the absence of one considerably more expensive.
Yes, and the case against it is usually a case against a particular execution rather than the channel. The harder truth in this conversation is that a cold call now gets roughly fifteen seconds to land, so the constraint is the quality and consistency of the opening message rather than the number of dials. An MSP getting poor results from calling should audit what is actually being said on those calls before it changes channels.
On price alone, often yes. The example in this episode weighed roughly fifty to sixty thousand dollars a year for an appointment setter against thirty-six thousand for an AI outreach service. That comparison is only meaningful if both are measured against the same outcome, such as first-time appointments booked. Compared on opens and clicks, the two are not being compared at all.
State the outcome first, then make them answer against it. A usable version: my goal is ten first-time appointments a month, this is my ideal client profile, these are the qualifications, what can you do to help me reach that goal? The answer reveals whether the vendor is accountable to a result or to an activity, and it does not require you to understand their technology.
Probably not, and expecting it to is the wrong reason to buy. Recipients are running increasingly capable filtering on their own side, so deliverability is close to a wash. Where AI genuinely changes the economics is upstream: sharper targeting, list building from your own network and its connections rather than a purchased list, and materially better message quality and follow-up.
Bering McKinley's rule is thirty days in the deciding stage. The specific number matters less than having one. A company that is going to work with you generally decides inside a month, and holding older deals open inflates the pipeline with opportunities that are not real. The contact stays and moves to nurture. The deal closes.
The message, and a way to check it. Confirm that the owner, the salespeople and anyone setting appointments describe the company the same way, then build a means of assessing calls and emails against that baseline on an ongoing basis. Outsourcing the message and the channel together, before either has been tested, is the most common way this spend disappears.
Gary Boyle is a Partner for Strategy and Business Development at Bering McKinley. With a background spanning network engineering, entrepreneurship, and strategic consulting, Gary brings real-world operator experience to helping MSP owners build stronger, more profitable businesses.
Connect with Gary on LinkedIn →
Josh Peterson is the CEO of Bering McKinley and host of The BMK Vision Podcast. Since 2004, Josh has worked with hundreds of MSP owners to build operationally sound, profitable businesses through consulting, peer teams, and direct coaching.
Connect with Josh Peterson on LinkedIn →
Naming the outcome you are buying, and knowing whether your message can survive being amplified, are both decisions that belong upstream of any channel spend. That sequencing is what the Vision Operating System is built to enforce.