MSPs Are Drifting Toward an AI Pricing Niagara
Many are sticking with per-user fees doomed to decline over time. Plus: What partnering with Anthropic’s like and ThreatCaptain’s dollars-and-cents approach to selling security.
Where there’s mystery, a familiar saying goes, there’s margin. By that logic, there’s a whole lot of margin to be had in AI pricing. The question, however, is where and it’s an important one.
“The companies who figure out pricing and packaging the fastest will have a big edge in the early days of this AI phase shift,” wrote Altimeter Capital partner Jamin Ball earlier this year. “Price too low and you’re literally paying customers to use your product. Price too high and you lose to the competitor who figured it out first.”
He’s not the only SaaS investor pondering the topic either, nor are SaaS investors alone in thinking about it. SaaS companies are sorting through it as well. ServiceNow, for example, is transitioning from a traditional per-seat pricing scheme to a hybrid one combining a per-user subscription fee with consumption-based access to AI functionality.
There are plenty more examples where that one comes from, too, as vendors adjust to the reality that agentic functionality lets businesses get more done with fewer end user licenses while imposing unpredictable, potentially steep upstream LLM costs.
Nor does the hunt for new AI-native pricing strategies end with software makers. AI consultants, like the one we introduced you to here last week, are wrestling with the issue as well.
“People are experimenting with pricing all over the board,” says John Bush of Dallas-based consultancy wave4, who once priced a relatively simple client project involving an automated quoting and estimating tool at a flat $7,500. “He goes and talks to other vendors for that same use case and gets pitched a $10,000 a month minimum contract, and you have to sign for a year.”
From consultants to software companies to VCs, everyone, it seems, is searching for the right approach to AI pricing.
Except…
“Everybody that I talk to seems to be looking at this, except for maybe all the MSPs, who maybe are not,” says Omdia principal analyst Jessica C. Davis (pictured).
Well, some are. Per the results of a recent Omdia poll sent to 22,000 partners worldwide, 43% of MSPs are evaluating an alternative to user-based pricing and another 17% know they need to make a change but haven’t decided what kind. A full 40%, however, aren’t even thinking about modifying the per-user, per-month rates they’ve been charging for years. Davis, paraphrasing her Omdia colleague (and previous Channelholic interviewee) Robin Ody, likens that last disturbingly large cohort to a bunch of sightseers drifting toward Niagara.
“Our boat is headed towards the big waterfall and we’re going to go over the edge, but nobody’s figuring out what to do about it yet,” she says.
Which means, assuming Omdia’s numbers are right, that roughly four out of every 10 MSPs on Earth are about to get seriously wet. Say your average client has 35 employees suitable for per-user pricing at present. “That number will go down on average,” says Colin Knox, previously an MSP himself and currently CEO of billing management vendor Gradient MSP, until eventually your younger clients have 10, five, or even just one user to support.
It may not take as long as you think to get there, either. As you read here in June, Elliott Hyman, CEO of mega mega MSP Lyra Technology Group, is already serving “healthy growing businesses with declining headcount.”
Fewer users is just half of the problem too. Many MSPs, Knox observes on LinkedIn, base their per-user rates on how many hours they spend supporting a typical user. Multiply that number by what each hour costs you, add in some margin, and multiply the result by the number of users you support, and you end up with an MRR total. Except that AI reduces both the number and cost of the hours, by automating the help desk, and the number of users, by increasing customer productivity.
Hear that waterfall up ahead? It’s what researcher Luke Ronkowski of TSIA calls the MSP profit paradox: “the more value your AI delivers, the less you earn.”
Imperfect options
Alright, so maybe embracing a pricing model based on the tokens your AI consumes is the answer. A little over 30% of Omdia’s poll sample plans to do exactly that, either exclusively or in combination with a per-user base charge. Sounds simple, but it isn’t. For one thing, good luck forecasting what the AI will cost you.
“Token prices are falling fast, hardware efficiency improves every generation, and the mix of models you’re calling changes the math entirely,” Ball observes. Fix your rates at a number that doesn’t change even after OpenAI slashes prices on a model by 80% overnight and your growing margins will have trouble keeping up with your shrinking revenue as underpricing competitors lure your customers away.
“Really, I mean, it’s almost not worth reselling tokens,” says Duane Barnes (pictured), president of global managed cloud service provider RapidScale. “It’s very hard to make money.”
And risky to boot. “If the customer pays you late and you can’t pay Anthropic late, you’ve got a problem,” Barnes notes, which is why RapidScale credit checks customers before approving them for usage-based token pricing and has little patience for deadbeats.
“We’ll disconnect the account and pause all consumption while we wait for payment,” Barnes says. “And for good reason. Those dollars get big very fast.”
Better then, perhaps, to avoid per-user fees and per-token fees in favor of outcome-based pricing of the kind Fin, the customer service vendor formerly named Intercom, helped make famous in the years before Salesforce bought it in June. Most businesses, after all, like the sound of paying only for actual value versus potential value.
Except that too is simpler than it sounds, for reasons I explained a long while ago. Call center outcomes are easy to monetize. If the AI resolves your customer’s complaint, you pay. If it doesn’t, you don’t.
Managed service outcomes are harder. Will clients willingly pay you every time you close a ticket? Even if they don’t understand what the tickets are for? Or agree that you fixed the issue to their satisfaction every single time you close one? Besides, service desk automation software may be turning tickets into an endangered species, and alternative outcomes are even harder to track, validate, and bill for.
Plus, isn’t outcome-based pricing exactly the kind of non-recurring revenue model people are escaping when they become MSPs? “It almost sounds like break-fix pricing again to me,” Davis says. “It takes away the predictability of revenue for the MSP depending on how many users are having problems in a particular month versus not.”
Barnes prefers what he calls an “Accenture-esque model” in which RapidScale gets paid a percentage of what an AI solution saves its users. Here too, though, it helps to have experience most MSPs lack.
“We’ve done enough of these, many hundreds of them over the years, that we have a feel based on the type of company, the industry they’re in, and other companies we’ve done similar work with for roughly how much we think we can save them,” Barnes says. The price it derives from that estimate, though not designed to be outcome-based, arguably is.
“Did we accomplish the thing we jointly set out to accomplish with your teams? Yes or no?” Barnes asks. “If no, let’s figure out how it went wrong. If yes, great. Sign off and let’s work on the next project.”
Of course, most MSPs haven’t delivered hundreds of AI solutions and don’t yet know what kind of savings if any the solutions they’re starting to build now will provide. Which is why Davis advises them to experiment with pricing models while they gain experience and brace themselves for potentially expensive bumps and bruises in the months ahead.
“It’s going to be a learning time, and there’s going to be pain involved in it,” she says.
Over on The Business of Tech
Host Dave Sobel recently spoke with Davis about AI pricing and other interesting topics at length. Check it out here and enjoy.
A direct line to Anthropic
One more note about RapidScale. In addition to being a managed cloud service provider with an Accenture-esque AI solution methodology, it’s also an Anthropic partner.
Which is an interesting thing to be mere months after the $100 million launch of the Claude Partner Network, and several additional months after Anthropic introduced MSPs to a whole new form of AI-related risk they need to manage.
That was when the AI giant (“for, I’m sure, very good reasons,” Barnes says) gave businesses transacting on the AWS Marketplace through an MSP 60 days to either cut the MSP out of that loop or replace them with an authorized reseller, a status RapidScale didn’t yet hold at the time.
“We had to act fast as we had a number of customers consuming Anthropic through the marketplace,” Barnes says. “They would have been shut off if we didn’t act, or at least advise them to go get a direct relationship with Anthropic.”
Instead, RapidScale joined longtime alliance partner AWS as a Claude Partner Network member. The specific benefits it’s been enjoying ever since are under NDA, but include preferred pricing, MDF eligibility, and early access to cutting-edge models.
“If you’re on the 5.1 model, we’re on the 5.3 model, typically a little bit ahead of everybody else,” Barnes says. The biggest benefit of all though, he continues, is the ability his team now enjoys to speak directly with Anthropic engineers.
“When we need help, when we have crazy ideas we want to bounce off them for trying new things, they’re very responsive,” Barnes says.
And “the best of the best in our experience,” he adds. That makes having a direct line to them a pretty serious advantage, obviously, as the folks at Shield Technology Partners (who have direct access to OpenAI engineers) can attest. That said, Barnes insists, no one should let the fact that they’re not on a short list of Anthropic service partners prevent them from building Claude into their AI practice.
“The customer can go pay their Anthropic bill directly,” he notes, while you make money developing, supporting, and securing Claude-based solutions, not to mention writing acceptable use policies for them and teaching people how to use them.
“There are a lot of things you can add value and do without ever reselling the license or Anthropic even knowing who you are as an MSP,” Barnes says.
And besides, if they don’t get those things from you, Anthropic itself might just step in to do the job for you.
Speaking of AI…
Remember IgnitHQ, the upstart, AI-native tool suite for MSPs I wrote about a few months back? Here’s your chance to hear its whole, very interesting story directly from founder and CEO Yaron Baitch on MSP Chat, the podcast I co-host. That episode is available here. Other, equally interesting episodes can be found here.
The problem isn’t what you sell, according to ThreatCaptain. It’s what you say.
Let’s put a great big generalization out there. Most MSPs are very good at service delivery, pretty good at security service delivery, OK at sales, and terrible at security sales.
Easy enough to see for yourself if you don’t believe that last part. Just ask the next 10 MSPs you speak with how many have never had a client refuse to buy even their minimum viable security service package. If you find one, I’d love to meet them.
According to ThreatCaptain, a security startup founded a little under three years ago, the issue isn’t that MSPs don’t know how to sell so much as that they don’t know what to say in a sales meeting that will actually break through with business owners.
“Typically what ends up happening is an MSP will explain what needs to be done, what the dangers are, and what the risks are, and then the client says, ‘So what?’” explains ThreatCaptain co-founder Adam Anderson (pictured).
Not because they don’t believe phishing and ransomware and autonomous agents skilled at breaking and entering exist, but because such things exist only as abstractions to cash-conscious decision makers.
ThreatCaptain’s software aims to make that abstraction concrete. Drawing on data breach cost statistics from sources like Verizon’s annual DBIR, frameworks like CIS and MITRE ATT&CK, and thousands of Monte Carlo simulations, the system (which was originally designed to help insurers model cyber risk) calculates how much financial damage a client would suffer after a successful attack. That, in turn, positions MSPs to have what Anderson calls a “normal, real” discussion with customers in plain, non-technical language about how much exposure they can stomach and what it would cost to stomach less.
“They have the leadership of their clients saying, ‘Well, I understand this is a $50,000 a month risk I’m carrying. I think I could probably deploy $15,000 a month to lower that. What can you do to help me?’” Anderson says. “Completely different conversation.”
That functionality, called First Mate and originally the whole of what ThreatCaptain offered, is now one of three modules in an expanded version of the solution. The second, named Anchor, helps MSPs find and research sales prospects. The third, called Admiral, is a “risk registry” system that draws on real-time threat intelligence to identify sales opportunities among existing customers. Say a report crosses the wires indicating that ransomware attacks on dentists are up 20% this year.
“Well, then we automatically have the ThreatCaptain risk registry look at the MSP’s customer base and identify all the dentists,” Anderson says.
Collectively, all three components form a platform offering “an all-in-one sales methodology backed up by software that guides MSPs, their salespeople, their sales engineers, and the consultants through that entire customer journey,” Anderson says.
First Mate and Anchor both enter general availability on August 3rd. The former sells at $199 for 50 enriched leads a month, with more available at extra cost. The latter costs $599 a month. Admiral, due out in September, covers 10 clients for $399 a month, and users can add more clients on a pay as you go basis. Those are all easy sums to justify in Anderson’s view.
“The net impact of a platform that guides the conversation away from tech jargon and back to business outcomes and that is focused on salespeople and customer success people is that your MSP is going to sell more cybersecurity quicker,” he says. More revenue for you. Less risk for your clients.
Also worth noting
Interesting: Acronis now offers a unified MSP workspace for protection, automation, and infrastructure, plus a new AI-powered autonomous service desk tool.
Veeam Data Platform v13.1 offers broader hypervisor support, expanded malware scanning and cloud threat detection, and more.
1Password’s newly launched new Privileged Access tools is designed to eliminate standing privileges across human and AI identities.
SonicWall is the latest security vendor with access to Claude Mythos 5 now that it’s been admitted to Anthropic’s Project Glasswing.
Fortinet’s new FortiGate 1200G and FortiSASE Outpost aim to combine high-performance, ASIC-powered firewall protection with cloud-delivered SASE capabilities.
Torq has shipped SOC Brain, an AI capability designed to give security operations teams contextual reasoning and autonomous decision-making.
Abnormal AI has added identity threat protection, AI governance, and infiltration prevention to its Behavioral Security Platform.
KnowBe4’s new simulated vishing capabilities are designed to help employees spot increasingly sophisticated voice-based social engineering attacks.
Red Sift has added protection for Claude and ChatGPT Cowork to its email security platform.
Checkmarx’s new Fusion Hybrid Scanning capability combines multiple application security testing methods to deliver more comprehensive vulnerability detection.
Keyfactor is buying Cofide to add verified identities for AI agents and cloud workloads to its platform.
BlackCloak’s deepfake protection now covers an executive user’s family members, advisors, assistants, caregivers, and other close contacts.
The Qualys TotalAI platform can now discover, monitor, govern, and secure AI models, agents, MCP servers, and shadow AI deployments.
Could be wrong, but I believe ThreatLocker’s a mere $11 million shy of a half billion in lifetime funding following the close of its new $190 million Series F round.
80% of breaches began with stolen identity in the first half of this year, but 27% involved SSL-VPN appliances, according to Cynet.
Dell Private Cloud’s PowerStore storage now integrates with Nutanix Cloud Platform to deliver enterprise storage performance, automated deployment, and cyber resilience.
Intune for MSP partner inforcer has closed a $50 million Series C funding round.
Speaking of Intune, 81% of IT pros are thumbs up on Microsoft Intune yet 65% spend 6-15 hours a week on manual packaging, patching, and reporting, according to Recast.
Mega mega MSP Omega Systems has a new CRO, COO, and CFO, plus three new VPs.
Chris Malone is the new CEO and Conor Burns is the new CFO at Sumo Logic.