Strategic Sales Teardown: Perplexity

A member of The Purposeful Performer wrote to me a while back in our community with a request. He saw the strategic selling teardown I did on Klaviyo and asked if I could do the same for Perplexity.

Good timing. Perplexity is hiring for an Enterprise Account Executive right now. The role is full-cycle and focused on building "senior-level relationships across enterprise clients,” with $200,000 to $400,000 in on-target earnings. In March, their Chief Business Officer, Dmitry Shevelenko, told VentureBeat the company had "tens of thousands of enterprise customers and only six people on our enterprise go-to-market team." Now they're staffing up.

I found this interesting for a reason unrelated to Perplexity. An AI-native company is hiring human sellers with a potential $400K OTE. Whatever you've read about AI coming for sales (or the end of humanity, for that matter), the market is planning for something different.

So I did the exercise the way I'd do it if it were my job. I “got hired” for the Austin geography and was assigned a strategic account to pursue in the market—CrowdStrike. And I gave myself nine months to close a seven-figure strategic transformation deal.

They're already a Perplexity partner. More on that in a minute.

CrowdStrike sells cybersecurity software and is headquartered in Austin. They have about 10,700 employees, $5.84 billion in annual recurring revenue as of July, and are growing 25% a year. Their CEO, George Kurtz, told investors in August that the quarter was "the best quarter in CrowdStrike's history."

In March, CrowdStrike and Perplexity announced a partnership. CrowdStrike's security platform, Falcon, would protect Perplexity's browser for enterprise customers. The press release focuses on protecting other companies that use Perplexity. It says nothing about CrowdStrike using Perplexity itself.

That gap is the deal I want to pursue in this exercise, and unpacking the approach may spur ideas for setting up your own strategic pursuit.

The first place I’d start is by reading everything George Kurtz and Aravind Srinivas have said in public over the last year. Using nothing but Perplexity (for this exercise though, I actually used Claude Fable 5.1 because that is my go-to LLM), I sourced earnings calls, keynotes, interviews, and public-facing posts and essays.

I wanted to know what they'd talk about if they were in a room together. It turns out they're describing the same company from two perspectives.

Kurtz said this ten days ago:

And on what his customers want:

His customers' words to him, in his own words: "We can't go fast enough in the business."

His ideal end state is a level of automation "where the car drives itself," with "the customers going to control the level of automation they have or want and where they want it."

Last year, he published a five-level scale for exactly that, for his security operations center, the team that watches for attacks around the clock. And on his August call, he said every employee will soon have about 90 AI agents working for them.

Now, here’s what Perplexity’s CEO, Srinivas, has been saying since February:

He went on to say what holds it back inside big companies: "A continuous investment in security engineering is necessary." And his one prediction for the industry, from a talk last November: "accuracy. Everything else breaks without it."

Kurtz, on the same subject, to analysts: "In cybersecurity, you simply cannot have a hallucination. It is first time, final."

Now let’s put these side by side. Kurtz has the control plane and the customer's fear. Srinivas has the computer and the customer's ambition. Neither one has a large company that has put the two together and said so in public.

Every big deal I ever closed had a category behind it. At LivePerson we called it going conversational. Everyone else was talking about omnichannel customer experience. We went to market with a different point of view and said the whole way a brand talks to its customers should be focused on messaging, not everywhere at once, and then showed them the old world and the new world, backed by real large-scale deployments who were led by innovative leaders.

Here's the simple choice for this deal.

Become an Autonomous Company.

It’s a company where the work runs itself, and the people decide where and how far. It’s not "AI-assisted," where every employee gets a license and keeps doing the same job with a co-pilot. It’s fully autonomous, where whole pieces of work are handed to secure software that runs while the best people focus on high-value strategy and execution at the highest standards. Every piece is visible and under control, and each team picks its own level of autonomy (using Kurtz’s five-level framework above).

For instance, the sales team might decide it's at level three, meaning agents draft account plans and pre-call briefs on their own, with a person checking before anything goes out. Legal might decide it stays at level one, meaning AI is on company accounts and visible only to those with credentials, but humans still do all the work. Engineering might go to level four. In other words, it’s the same company, different settings, each chosen by the team that owns the work, because the people closest to the work know which parts are safe to hand off and which aren't.

Old world (the status quo organization): AI-assisted. Licenses handed out. Shadow AI tools dispersed everywhere. Nobody can name what's running. Speed held back by fear. No cohesion on accomplishing strategic goals.

New world (the modern, elite organization): Autonomous. Every team on a stated level, zero to five. Every agent is visible. The best people set the level and check the work. The company goes as fast as it decides to.

Why this one? It's Kurtz's own words. He has a Chief AI and Autonomous Systems Officer. He built a five-level autonomy scale for one department. It's also Srinivas's own claim. And it's the word I've been writing about for three years, so I can tell the story with some authority.

The line I'd put at the top of the internal memo: The SOC drives itself. Now the whole company can.

The first thing I'd do in the new job is not open the CRM or LinkedIn. I'd write the authoritative article—one written well enough that the Wall Street Journal would run it. It would define the category and how a top security company would run itself that way. No product in it. No vendor. Public facts only.

That article is the DNA of the whole deal. Every LinkedIn post, every email, every creative outreach, and every meeting pre-read gets cut from it. Here it is:

Ask a chief executive what's running inside their company right now, and they'll name the software they purchased last year. Ask the security chief, and you'll get a different answer: "We can't tell."

That's not a guess. George Kurtz, who runs CrowdStrike, the Austin-based cybersecurity company with $5.8 billion in annual recurring revenue, said it best last week: "What we see in the enterprise right now is a lot of adoption of AI, but also a lot of shadow AI. People can't figure out what's running." One recent industry study found that nearly half of all workplace AI conversations happen on personal accounts, invisible to the company.

Most boardrooms have responded by slowing down. Mr. Kurtz says his customers want the opposite. They "want to go faster with AI deployments, but they're being held back because they can't implement security, governance, and a control plane around these agents." Their message to him, in his words: "We can't go fast enough in the business."

Here is the strange part. The tools to go fast exist. So do the tools to stay safe. What doesn't exist yet is a large company that has put the two together and said so confidently in public.

What they're all circling around is called the autonomous company. Not "AI-assisted" or “AI-native” or “AI-powered,” where every employee gets a license and keeps doing the same job with a helper. The autonomous company means handing off whole pieces of work to secure software that runs while people focus on the most valuable work, where each piece is visible and under control, and where each team decides for itself how far to go.

That last part matters. Mr. Kurtz said it clearly: "The customers are going to control the level of automation they have or want and where they want it." He even published a scale for it, five levels of autonomy for a security operations center, from a human doing everything to a system that, as he says, "drives itself."

Five levels, one department. The obvious question is why stop there?

A security operations center is a few hundred people at most. A company like CrowdStrike employs about 10,700. Each one will soon run software agents on the company's behalf; Mr. Kurtz estimates the number at roughly 90 per person. Sales, finance, legal, engineering, support. If a five-level scale is the right way to govern autonomy in one building, it's the right way to govern it in the other fifty.

An autonomous company would publish its levels. Level one: every tool an employee uses runs on a company account, so nothing is invisible. Level two: the work that happens in the browser, which is where most knowledge work now happens, is protected the same way the laptop is. Level three: teams can hand off whole tasks, with a spending limit and a record of every action. Level four: the company's own security watches its own agents, and can show a customer the log. Level five: the playbook is public, and other companies follow it.

Why should a cybersecurity firm go first? Because it has the most to gain and the least excuse not to. Its entire pitch to the market is that agents can be secured. Its own CEO says securing them is "the largest market opportunity in our history." The fastest way to prove that claim is not another product launch. It's to be the first large company where every employee's agents run inside the company's own protection, and to say so on an earnings call.

There's a cost to going first. Somebody has to write the levels, pick the teams, and be accountable when level three doesn't work the first week. There's a bigger cost to going second. The first company to publish its autonomy levels gets to define what "autonomous" means for everyone who follows, the way the first company to say "we stop breaches" got to own that sentence.

The autonomous company is coming either way. The only question is whether a security company will show everyone else how, or become one more logo on the list of firms that waited until it was safe.

Given what its own CEO has been telling Wall Street, waiting would be the strange choice.

That's the article. It’s not meant to be a final artifact. It’s purposefully 70-80% done to make the job about finding and refining the extra 20-30%. Notice what it never does. It never says Perplexity. It never says "our product." It leads to the solution. It doesn't lead with it.

Most sellers would take this to the CISO. That's the wrong door. The CISO is who CrowdStrike sells to. They don’t buy inside CrowdStrike. They can stop the deal in month seven, and I'll get to them then.

The owner is the Chief AI and Autonomous Systems Officer. The category is in his title. He joined in May after serving eight years at NVIDIA. His new lab was announced two weeks ago. He needs a first win, and the article is written for him.

He's the most likely Mobilizer, or at the very least, the fastest path to one. A Mobilizer is not a champion. They’re the champion of all champions. This is the executive who cares more about the company than about their own department, and who will give you friction first because they’re checking whether this is actually good for the company.

We find out if he’s the one by finding a creative way to get the article in front of him. He needs to just answer one thing: "This article was written about your company. Tell me what I got wrong." But there’s a lot of work to do to get to that stage.

A good starting point is figuring out a path to the Chief Business Officer, who signed the Perplexity partnership in March and already said: "the browser is no longer just where people access information – it's becoming the interaction layer for enterprise AI." He's a door to the C-Suite. The CIO implements. The CFO shapes the money. The CISO checks.

Using only Claude and no special software, I mapped a v1 chart of the whole organization below the C-suite, down to VP level. It helped me identify 25 executives, in the order they enter the deal. Here's the map:

The goal is to get to a Mobilizer quickly using every avenue possible, whether it be leveraging the executive-level contacts via the strategic partnership already in place, an open letter sent digitally and in the mail, specific content on LinkedIn pointed at specific executives, or most likely, a combination of all of the above.

Once we establish contact and the Mobilizer has marked up the article, we assemble a cohesive team to build out a maturity model together. Every level uses CrowdStrike's own words: Falcon, Flex, Guardian, Charlotte. In plain words, the five levels are:

0. Ungoverned. Employees use whatever AI they find, on whatever account they have. The company sees the laptop, not the agent.

1. Everyone inside. Every employee's AI runs on a company account, one login, controlled by IT. Nothing runs in the dark. This is a 90-day project.

2. Protected on every desk. The AI browser is secured by the company's own platform. This is the partnership they already announced, run on themselves.

3. Agents for the people who need them. Engineering, threat research, sales. Whole tasks handed off, with a spending cap per person and a record of every step.

4. Prove it on ourselves. The company's own agent-security product watches the company's own workforce. The internal numbers become the customer story.

5. Teach every customer. CrowdStrike publishes the playbook. Every one of its customers gets the same path.

Nobody at this size pays list price for seats. So I wouldn’t sell seats.

Kurtz on the CrowdStrike August earnings call: "Flex is the commercial harness to enable customer success in the agentic era." His CFO said customers who move to Flex show a 40% increase in annual spend on average. So, I’d propose a commercial model aligned to this type of outcome.

The goal would be to start with a three-year commitment, justified by the strategic partnership already in place. Three years is a good starting point, allowing both sides to pivot to a two-year or one-year deal that protects either side during negotiations.

It would center on one number the company draws down against for people, agents, and usage as each team moves up a level. Year one is the seven-figure floor. The pitch: we'd like you to buy from us the way your customers buy from you.

It also gives two outcomes both sides can measure. First, the share of employee AI activity that runs inside company identity, with the target being all of it. That's shadow AI at zero, a number the CEO can say on stage. Second, the number of teams that reach level three, with labor value measured the way Perplexity already measures it. Hit both and Perplexity earns a bonus. Miss, and part of year two comes back.

One alignment clause. CrowdStrike becomes the first published autonomous company, and the two companies sell that path to CrowdStrike's customers as one offer, with a share back to CrowdStrike. Now the deal isn't a cost line for them. It's a product line. That's how one logo becomes a channel.

Most software gets sold one line item at a time. A company buys 500 seats for people. Then a separate contract for agents. Then a separate bill for usage. Three contracts, three budgets, three renewals, three fights with procurement.

I'd propose one contract instead.

CrowdStrike commits to a single dollar amount up front. Call it a baseline of $6 million over three years (using this as a broad number without having insider knowledge, but appropriate considering CrowdStrike’s size and Perplexity’s current Enterprise seat-based pricing).

They can spend it on anything Perplexity offers: seats for humans, agents that run on their own, or usage credits. They draw down against it the way you'd spend from a prepaid account. Every seat added, every agent deployed, every usage spike comes out of the same pool. They’d get favorable tiered pricing above that baseline if they need more

Here's where it connects to the five levels.

At level one, a team mostly needs seats, because people are still doing the work with AI beside them. At level three, that same team needs fewer seats and more agents, which would run at a lower cost, because the agents are doing the work. At level five, it's mostly usage, because there's almost nobody in the loop.

The mix changes as a team climbs. The contract doesn't. The commercial model incentivizes them to reach full autonomy as quickly as possible, without cutting corners. It would be paired with a custom strategic customer support plan tailored to CrowdStrike that ensures quality assurance and executive alignment. It would be priced as an annual subscription, adding an additional six figures in ARR to the work order. Nobody renegotiates every time a team gets more autonomous. They just spend the same pool differently.

This removes the biggest reason enterprise AI deals stall. The buyer doesn't have to guess in month one how many seats versus agents they'll need in month twenty. They commit to the number, and the mix sorts itself out.

And CrowdStrike already sells this way. Falcon Flex is exactly this structure for their own customers. So when this lands on Burt Podbere's desk (the CFO), it won't look like a vendor's clever idea; it'll look like their own enterprise customer playbook and commercial model.

Most account plans are a list of things to do in a deck. I’d build one cohesive motion using my calendar like a game console.

The deal will be won, lost, or delayed in how the days are shaped, because the plan only works if the person running it has the energy to think clearly on a Tuesday in month six. So this is the whole operating system—the work and the human running it:

5:45 a.m. Wake, no phone. Hydration, meditation, and mental preparation. The phone stays in sleep mode and in another room until the first deep-work block is done. The best thinking on this account happens before anyone else's agenda arrives.

6:15 to 7:00. Movement and light. Lift, ride, or run, then shower. Not necessarily to get fit, but to get the mind ready to do the next block. On travel days, it's a 20-minute walk outside. It never gets skipped, because the days it gets skipped are the days the deep work gets skipped.

7:00 to 7:30. Fuel and read one valuable thing. Fuel up. Then one earnings call transcript, one CrowdStrike blog post, or something meaningful for the article. The high standard I would hold myself to: Know enough about them that I could get hired by them. 15-20 minutes of input before any output.

7:30 to 9:30. Deep work on the account. This block, done consistently, ultimately helps close the deal. I shut the door, turn off all notifications, no Slack, no email, and just one task: researching and writing the article, the outreach, the org chart, the memo, the maturity model, the business case. Nobody else gets this time. Not my manager, not a "quick sync." If someone needs me, they can reach me at 1:00. I commit to 3 Pomodoros of 25-minutes each, which works out to about 90 minutes of work and about 30 minutes of breaks.

9:30 to 12:00. Internal work. Perplexity's partner team, product, security engineering, or ideally, the Chief Business Officer’s office. I’d go deep with the people inside my own company who manage the CrowdStrike relationship. Most sellers skip this and go straight to the prospect. The prospect can wait. The inside details that provide a gold mine of insights can't.

12:00 to 1:00. Lunch away from the desk. A walk after. No account thinking. The afternoon needs a different brain than the morning.

1:00 to 4:30. Conversations. Calls, outreach, follow-ups, the Mobilizer, the field. People work is afternoon work, because by then I've already done the thinking that makes the conversations worth having.

4:30 to 5:00. Close the day. Update the account notes. Write tomorrow's one key move. Clear the inbox to zero or to "Monday." Write down one thing I’m grateful for. Then shut the laptop and don't open it again.

Evening. Dinner with my wife. No screens after 9:00. In bed by 10:00. 7-8 hours of quality, consistent sleep is not a goal; it's a requirement for quality performance.

Monday, 8:00 to 9:30: the weekly organization block. Before anything else, 90 minutes alone with three questions.

This block is scheduled in advance every week and treated as sacred; otherwise, it’s too easy to skip or get pushed aside by someone else’s agenda. This block replaces the 7:30 deep-work block on Mondays. It's the most important 90 minutes of the week.

Tuesday and Thursday mornings: the two big deep-work days. The 7:30 block runs long, to 10:00. These are the strategic blocks for finding justifiable inroads into the account.

Wednesday: the internal day. This is time with the product team, partner team, and security. It’s about building the team on my side of the table so that when the two teams meet in month four, we are ready to move with speed and confidence.

Friday, 3:30 to 4:00: one edit to the narrative. The executive memo that eventually transforms into the narrative business case is a living document. It gets one edit a week, based on what was uncovered in that week. It will be the artifact at the center of everything we use to keep our deal progressing forward. Then the week closes early. No calls with the customer after 4:00 p.m. on Friday, unless the client absolutely requires it.

Saturday: solo bike ride. Sunday: 30 minutes of reading and a long walk with Lisi. The account does not exist on the weekend. The people who burn out on nine-month deal sprints are the ones who carry work into the weekends.

First Monday of every month, 3 hours: the strategic session with myself. Effort versus impact on everything I did in the last 30 days. Two columns. What actually moved the needle on the deal. Then three decisions: what to stop, what to hand off, what to do more of. This is where the plan gets corrected and where any blind spots get worked through.

For the rest of the day. A ride, a hike, an afternoon with no laptop. It’s scheduled on the calendar in advance, with no exceptions to cancel it. The deal lasts nine months. Nine months of full gas commitment without a full strategic recovery time built in is a recipe for burnout or unnecessary mistakes.

Before every executive conversation, 30 minutes with the internal team and then 10 minutes alone. Three questions, written: What do I want them to feel when this ends? What's the one thing I know they don't? What's the one thing they might say that would rattle me, and what's my answer? Quiet my mind and run through some controlled, slow breathing. Then the call or head to the meeting.

After every executive conversation, 45 minutes. What did they actually say? Not what I wanted to hear. Verbatim, in the recorded notes, before it gets lost. Debrief with the team after an executive meeting, and commit to assigning a next step and owner.

The standard I hold myself to. No email, memo, or slide goes to CrowdStrike until it gets a second pair of eyes. We always ask, “If I were an employee at CrowdStrike in their role, would I respond to this?”

What I say no to. Internal meetings that don't move the deal forward or the two others on my list. "Pick your brain" coffees. Any recurring call without an agenda.

Month 1: Learn. Interview everyone inside Perplexity who touched the partnership. Read four earnings calls. Write the article. No outreach. Milestone: a one-page insight piece with no mention of Perplexity's product.

Month 2: Find the Mobilizer. Article to two or three candidates through the partnership relationship. Ask what's wrong with it. Milestone: one person who argued with the piece and then did their own research.

Month 3: Co-write the memo. The article becomes an internal executive memo, 60 to 80% drafted. Give the Mobilizer the chance to make revisions. Milestone: a version with their edits in it.

Months 4–5: Solidify alignment and define the workstreams. Executive working session through the maturity model. Build the five levels in their words. Identify the workstreams needed to bring it to life—Business, Operations, and Technology (including Infosec). Milestone: a multi-year maturity model both sides have signed off on.

Month 6: Business case. The memo becomes a narrative business case with the Mobilizer's name on it, not mine. Milestone: the CFO has an internal-facing document written by their own people using their own ROI analysis and taps procurement.

Months 7–8: Proposal and structure. Solidify the commercial model, and finalize security and legal redlines. The Comet vulnerability history gets raised here, and the answer is ready before the question. Milestone: redlines, not objections.

Month 9: Close. Multi-year, full company at level one, levels two through five as the roadmap in the contract's own language. Milestone: a signature, and a joint story both companies want to tell.

A lot better than emailing the CIO about a more powerful search tool, right?

This was 90 minutes of my time. I used no software other than my Claude subscription, running Fable 5.1. I orchestrated using my past knowledge base and deal wins. Claude proposed. I refined.

Get the whole method below: what I fed it, the six prompts in order, the filled-in system strategic sales sheet, and how to run it inside one of your own Diamond Accounts.