Jared Waxman writes the GTM Engineer Pulse and runs GTM Engineer School. Their work explains how businesses can build AI context layers and practical agent workflows for sales and marketing teams.
They are most useful for practical explanations of AI context layers and agent workflows for GTM teams.
Best for
B2B SaaS company piloting GTM AI context layers and agent workflows.
Not for
Broad sponsored content for a general B2B software buying audience.
Key takeaways
An agent withheld 15 customer-email replies because it could not find the company standard.
A sales team used a context layer to help reps position a feature for different partners.
Name which context layer you are building: brand, canonical or deterministic.
Check first: Confirm audience details and whether a paid talk or advisory format exists.
Main ideas
The model is not the moat. Where the context layer lives is. · Name which context layer you are building: brand, canonical or deterministic · Builders and consumers are different people in the same company
Jared Waxman writes the GTM Engineer Pulse and runs GTM Engineer School.
Their posts explain how sales and marketing teams can use AI agents, what information those agents need, and how to choose practical problems to solve.
The examples range from an agent that withheld 15 customer-email replies because it could not find the company’s standard to a sales team that used a context layer to help reps position a feature for different partners.
Follow their work for concrete cases and questions about how AI changes the decisions and workflows of go-to-market teams.
Waxman connects technical choices to the work people actually need to do.
Rather than treating “context layer” as one universal solution, they share a three-part distinction—brand, governed canonical context, and systems of record—and argue that teams should identify which they need.
They also show the limits: a folder of files may work until a sales team needs an always-current feedback loop, and an agent cannot follow a rule that was never written down.
The newsletter broadens the view with examples of changing sales decisions, new tools, and human approval boundaries.
Jared Waxman's recurring argument is that the hard part of putting AI to work in go-to-market is not the model.
Read more
It is where a company's context lives, who can read it, whether the rules an agent needs were ever written down, and which decisions a person still approves. They test those claims in hour-long conversations with operators, report what the field is shipping in a weekly roundup, and run a school where people learn by building one real workflow for one real company rather than studying the tools.
The model is not the moat. Where the context layer lives is.
Waxman makes the same argument from two angles.
Read more
Models and agents improve on someone else's roadmap, so they cannot be a company's advantage; what matters is where the context that feeds them lives.
No current option is neutral: GitHub suits agents but not humans, Notion suits humans but not agents, and a folder of files works until a sales team is in the field every day with years of history behind it.
Then the feedback loop from every call has to stay on.
Waxman spent an hour with Amos Bar-Joseph of Swan AI, whose post describes going from zero to 200 customers in a year with three founders and nobody else on payroll—no SDRs, customer success team or support team.
The point they keep returning to is not models or agents, but where the context layer sits.
They end by asking readers what stack they use to give their AIs org-wide context.
Bar-Joseph also shared his screen for half the conversation.
Waxman arrived at the GTM Engineer School discussion expecting an argument about running the school from a folder of markdown files. They got a conditional yes: files work as a starting point, but once a field sales team is working daily with years of history, a folder cannot keep the feedback loop alive.
Builders and consumers are different people in the same company
Waxman takes a specific position on adoption: a salesperson is never moving into Claude Code.
Read more
The people who govern and build the context layer need engineering tools and a release process; everyone else needs a chat interface with the plumbing hidden. Treating them as one audience is what makes an AI rollout stall.
At WorkSpan, the campaigns team and revenue operations went in first. Everyone else got Claude chat with the plugin installed and never saw the plumbing underneath.
Pick one problem with a deadline instead of building a company brain
Waxman relays the instruction not to make the project "build a company brain."
Read more
Choose one problem with a deadline and start there; the tooling follows the problem, not the other way round.
As Waxman puts it, all positioning is a theory, and the sales team is out there getting punched in the face testing it every day.
A dated, real problem is the smallest unit that survives contact with a buyer.
WorkSpan loaded its whole sales kickoff into Octave because the representatives coming out of it needed to position a feature to a services partner one day and a cloud partner the next.
That was a specific audience with a specific date, not an open-ended knowledge programme.
Waxman reports the quote from Zach Vidibor, who spoke during the same hour.
Write the rule down, then decide where the human stays
Waxman's lead item argues that an agent's ceiling is documentation, not capability.
Read more
An agent that cannot find a written standard does nothing, and that can be the better outcome.
Once rules exist, someone still has to decide which actions a person approves, and the field is split on how.
One reading of the week's job postings found a company granting agents autonomy through their acceptance rate, while another kept a human on every send, arguing that the gate gets worse as the system gets better.
OpenAI's newest model, Astra, was handed 15 customer emails to answer.
It looked for the company's reply standard, could not find it and sent nothing.
The older model could not find it either and sent all 15 anyway.
Astra still won the benchmark, which Waxman presents as the point.
In the same issue, Danny Chepenko's team at Grip went from taking the meeting with anyone to walking away from any company that is not growing.
The biggest change over eighteen months was being able to act on a standard the team already had.
Salesforce and HubSpot opened their CRMs to agents in the same week: one shipped a prebuilt MCP server with 37 sales skills inside Claude, while the other's server is now writing, with one thing it still cannot touch.
The two patterns Waxman's posts describe point in opposite directions: autonomy earned through acceptance rates versus a person on every send. They also note Wade Foster publishing the losses from 657 real business workflows, not just the wins, and six open GTM engineering roles spanning $87,400 to $240,000 under one job title.
Learn GTM engineering by shipping one real workflow for one real company
Waxman's priority is learning this work by building against a real problem, not by studying tools.
Read more
They run GTM Engineer School from a folder of markdown files in a GitHub repository.
Cohort 5, starting 26 October, puts 50 students each on one real workflow for one real company, with a 30-minute call a week.
It is free, not consulting, and carries no promised deliverable; the median participant in Cohort 3 had 12 years of experience.
Waxman also runs an hour-long recorded conversation format and a weekly roundup, GTM Engineer Pulse, which mixes interviews with items sent in by readers.
One issue features a Slack bot built in Cursor by an account executive turned GTM engineer, so DualEntry's sales team could run Clay without that person in the loop; a sorting of the GTM Substacks worth reading, with one bucket for people starting from scratch; and a daily intent score across 110,000 government institutions. Cohort 5's stated workflows are an enrichment waterfall, signal-based outbound, an inbound router and a CRM cleanup.
This answers the question "what is the argument about AI context actually about?"
Read more
Start here for the framing before the tooling: it sets up the tradeoff the other posts work through.
Waxman spent an hour with Amos Bar-Joseph of Swan AI, who reports going from zero to 200 customers in a year with three founders and nobody else on payroll — no SDRs, no customer success team, no support team — and shared his screen for half the conversation.
Waxman writes that they keep returning not to models or agents, which they say "will just get better and they're not your moat," but to where a company's context layer lives.
Bar-Joseph's claim, as Waxman records it, is that every option on the table is broken in a specific way: GitHub is good for agents and bad for humans; Notion is good for humans and bad for agents.
The post closes by asking readers what stack they are using to give their AIs org-wide context.
For brands planning a campaign
For a B2B SaaS team weighing a context-layer project, circulate this before any tooling decision.
It turns a vague ambition into a choice with visible costs on both sides.
A talk or advisory session can open with the human-versus-agent tradeoff and use Swan AI's headcount as the concrete case, while presenting the vendor claim as a claim, not a settled result.
This answers "what would a working context layer for a marketing or sales org actually consist of?"
Read more
Start here for the distinctions: it is the most concrete of the four and gives you vocabulary for an internal meeting that same week.
Waxman runs GTM Engineer School from a folder of markdown files in a GitHub repository.
They expected an argument when they showed it to Sam Gong, who standardized WorkSpan's marketing org on Claude plus a context layer, and Zach Vidibor, who built Octave to keep that layer current.
Instead, Waxman got a conditional yes: files are fine until a sales team is in the field every day with years of history behind them.
Then the feedback loop from every call has to be always on, and a folder cannot do that.
Three points from the hour recur.
First, name which context layer you are building.
Sam runs three: a brand layer anyone in marketing can write to; a canonical, governed layer for product and GTM strategy with a release cycle; and a deterministic layer of systems of record answered through a semantic layer.
Most companies say "context layer" and mean one of these without knowing which.
Second, separate builders from consumers: a salesperson is never moving into Claude Code, so at WorkSpan the campaigns team and RevOps went in first, while everyone else got Claude chat with the plugin installed and never saw the plumbing.
Third, do not make the project "build a company brain."
In Vidibor's words, pick one problem with a deadline.
WorkSpan loaded its whole sales kickoff into Octave because reps coming out of it had to position a feature to a services partner one day and a cloud partner the next.
Waxman's summary: "all positioning is a theory.
The sales team is out there getting punched in the face testing it every day."
For brands planning a campaign
This is the natural spine for a talk or advisory session on GTM context layers: it offers a three-part taxonomy, a builders-and-consumers split, and a scoping rule that steers teams away from an open-ended platform project toward one problem with a deadline. The kickoff example gives buyers a short, low-risk pilot they can copy.
This answers "what actually limits what an agent can do?"
Read more
Start here: one fifteen-email example makes the abstract point about documentation land in about a paragraph.
Waxman describes OpenAI's newest model, Astra, being handed 15 customer emails to answer.
It looked for the company's reply standard, could not find it, and sent nothing.
The older model could not find it either, but sent all 15 anyway.
Astra still won the benchmark, and Waxman writes that this is how it won.
Their conclusion: the ceiling for agents has moved.
It now depends on whether the rule an agent looks for was ever written down.
The rest of the issue rounds up Wade Foster running Astra through 657 real business workflows and publishing the losses as well as the wins; Adam Schoenfeld counting 34 AI companies past $100M ARR since January 2025, two of them in sales and marketing; a Clay Cup semifinal format giving competitors 25 minutes to clean a CRM, find the ICP, score accounts and draft outbound with no prebuilt workflows and no LLM; Christian Levan scoring 110,000 government institutions on intent daily; and six open GTM engineering roles ranging from $87,400 to $240,000 under the same job title.
For brands planning a campaign
Use this to open when an audience needs convincing that agent capability is not the bottleneck. It also gives buyers two reusable data points — a salary range for the same job title and a published-losses benchmark method — for a hiring or evaluation conversation.
This answers "what does the day-to-day of GTM engineering look like right now?"
Read more
Start here for the widest single snapshot of the field, from vendor releases to one person's Slack bot.
Waxman reports that eighteen months ago the team at Grip took a meeting with anyone; today they walk away from any company that is not growing.
They say that is the biggest thing the team can do now that it could not then.
The issue covers why it took 18 months to act and where the line now sits between what Grip's agents do and what a person still approves.
The rest of the round-up: Salesforce and HubSpot opened their CRMs to agents in the same week, one shipping a prebuilt MCP server with 37 sales skills inside Claude and the other's MCP server now writing, with one thing it still cannot touch; Joe Rhew read the week's GTM engineer job postings and found the job moving up a layer, with one company granting agents autonomy based on their acceptance rate and another keeping a human on every send; an account executive turned GTM engineer built a Slack bot in Cursor so DualEntry's sales team could run Clay without them in the loop; and a sorted list of GTM Substacks worth reading, including one bucket for people starting from scratch.
The issue announces Cohort 5 starting 26 October, in which 50 students each build one real GTM workflow for one real company — an enrichment waterfall, signal-based outbound, an inbound router, a CRM cleanup — free, described as not consulting and with no promised deliverable, on one 30-minute call a week.
For brands planning a campaign
The Grip example is the clearest illustration of an agent changing a commercial decision rather than a task, making it a better story for a sales audience than a tool tour.
The issue also shows how Waxman recruits for the school through a public capstone call rather than a paid offer.
For a partnership discussion, ask about the format of the work rather than assuming a sponsorship arrangement.
A weekly operator roundup and a cohort course built on real problems4
The context layer, not the model, is where the competition sits2
A three-layer map of what a company means by "context layer"1
Builders and consumers are different audiences inside the same company1
Pick one problem with a deadline instead of building a company brain1
Where the line sits between what an agent does and what a person approves1
An agent's ceiling depends on whether the rule was ever written down1
The context layer, not the model, is where the competition sits
Two of the four posts make the same argument from different angles.
Read more
In the Swan AI conversation, Waxman relays Amos Bar-Joseph's claim that models and agents will keep improving and are therefore not a moat: the battle is over where a company's context layer lives.
The problem, as described, is that every option on offer is broken in a specific way: GitHub is good for agents and bad for humans; Notion is good for humans and bad for agents.
The GTM Engineer School discussion sharpens the same point: a folder of markdown files works until a sales team is in the field every day and years of history accumulate.
Then the feedback loop from every call has to be always on, and a folder can't deliver that.
For brands planning a campaign
A talk or advisory session should open by asking where the client's context lives and who can read it, not by comparing models or vendors. That's the question Waxman puts to the audience, giving a workshop a concrete diagnostic first move.
A three-layer map of what a company means by "context layer"
Waxman reports that Sam Gong, who standardized WorkSpan's marketing org on Claude plus a context layer, uses three layers rather than one.
Read more
Brand is the layer anyone in marketing can write to.
Canonical is the governed layer for product and GTM strategy, with a release cycle.
Deterministic covers the systems of record, answered through a semantic layer.
Waxman observes that most companies say "context layer" and mean one of these three without knowing which.
Classification exercise: sort existing content and systems into brand, canonical, and deterministic, then check which layer has an owner and a release cycle.
The three layers are Sam Gong's setup at WorkSpan as described in the post, not a published industry standard, so a session should present them as one workable model rather than as a rule.
For brands planning a campaign
This gives a workshop a ready-made sorting exercise: ask a client to place existing material into brand, canonical, and deterministic buckets, then look for the missing governing layer and its release cadence. It turns a vague brief into a short list of build decisions.
Builders and consumers are different audiences inside the same company
Waxman takes a specific position on adoption: a salesperson is never moving into Claude Code.
Read more
At WorkSpan, the campaigns team and RevOps went in first; everyone else got Claude chat with the plugin installed and never saw the plumbing. This is presented as the workable division between people who build the context layer and people who use it.
For brands planning a campaign
An adoption plan should start with campaigns and revenue operations, then give the rest of the org a chat interface with the plumbing hidden, rather than training the whole company on engineering tools. For a client deciding who to enrol in a pilot, this identifies the two functions to start with.
Pick one problem with a deadline instead of building a company brain
Waxman quotes Zach Vidibor directly: don't make the project "build a company brain."
Read more
WorkSpan loaded its whole sales kickoff into Octave because reps coming out of it needed to position a feature to a services partner one day and a cloud partner the next—a dated, specific need. Waxman's own line from the same hour is that all positioning is a theory and the sales team is out there getting punched in the face testing it every day.
Waxman expected an argument from Sam Gong and Zach Vidibor and got a conditional yes instead: files are fine until a field sales team and years of history make an always-on feedback loop necessary. The approach is a starting point with a stated expiry condition, not a permanent architecture.
For brands planning a campaign
This is a scoping rule a client can act on immediately: name the single problem and the deadline before choosing tools, because the tooling follows the problem. It also sets expectations for an advisory session: the deliverable is a scoped pilot, not an enterprise-wide knowledge programme.
Where the line sits between what an agent does and what a person approves
In the GTM Engineer Pulse issue, Waxman reports that Danny Chepenko's team went from taking meetings with anyone to walking away from any company that isn't growing, and that the biggest change over eighteen months was being able to act on that.
Read more
The same issue covers Salesforce and HubSpot opening their CRMs to agents in the same week: one shipped a prebuilt MCP server with 37 sales skills inside Claude, while the other's MCP server is now writing.
It also features Joe Rhew's review of the week's GTM engineer postings, where one company's agents earn autonomy through their acceptance rate while another keeps a human on every send.
Rhew argues that the latter gate gets worse as the system gets better.
Autonomy gating: define which agent actions need human approval, then revisit the gate as accuracy changes.
For brands planning a campaign
For a client designing agent workflows, the approval threshold should be an explicit design decision with a named approver, not a default inherited from the tool. The two patterns in the postings—autonomy earned through acceptance rates versus a human on every send—give a team a useful choice to make early.
An agent's ceiling depends on whether the rule was ever written down
Waxman's lead item reports that OpenAI's newest model, Astra, was given 15 customer emails to answer, went looking for the company's reply standard, couldn't find it, and sent nothing—while the older model sent all 15.
Read more
Astra still won the benchmark, and Waxman concludes that the ceiling now depends on whether the rule an agent looks for was ever written down. The same issue notes Wade Foster publishing the losses as well as the wins from 657 real business workflows; Adam Schoenfeld's count of 34 AI companies passing $100M ARR since January 2025, with two in sales and marketing; and six open GTM engineering roles spanning $87,400 to $240,000 under one job title.
For brands planning a campaign
An audit of unwritten reply and positioning standards is a small, concrete first deliverable a client can approve before any build. The salary spread across identical job titles also gives a client a factual starting point for weighing whether to hire or build the capability internally.
A weekly operator roundup and a cohort course built on real problems
Waxman's recurring formats appear across all four posts.
Read more
GTM Engineer Pulse is a weekly roundup mixing original interviews with reader-sourced items: the Salesforce and HubSpot MCP releases, Joe Rhew's read of the job postings, naufal nugroho building a Slack bot in Cursor so DualEntry's sales team could run Clay without him, and Yuliia Harbych's list of GTM Substacks worth reading.
The longer formats are hour-long recorded conversations, including sessions with Amos Bar-Joseph and with Sam Gong and Zach Vidibor, each linking to a full recording on YouTube.
GTM Engineer School is described as running on a folder of markdown files in a GitHub repo.
In Cohort 5, from 26 October, 50 students each build one real workflow for one real company—an enrichment waterfall, signal-based outbound, an inbound router, or a CRM cleanup—on a 30-minute call a week.
The course is free, not consulting, and promises no deliverable; Cohort 3 had a median 12 years of experience.
For brands planning a campaign
A brand can attach to a recurring format rather than a one-off placement.
The structure sets expectations: one real problem, one real company, a short weekly call, and a working artefact at the end.
That is the clearest description in the posts of how a collaboration with Waxman would run.
Across the four LinkedIn posts, Waxman says they run GTM Engineer School.
Read more
The posts disclose no sponsorship, affiliate arrangement, or paid client engagement.
The companies mentioned — Swan AI, Anthropic's Claude, Clay, and OpenAI — appear as subjects of Waxman's interviews and newsletter coverage, not as disclosed commercial partners.
Ask Waxman for the posts' publication dates.
GTM Engineer School
Owned business
Jared Waxman says they run GTM Engineer School, which operates on a folder of markdown files in a GitHub repository.
Read more
Their newsletter describes a cohort in which 50 students each build one real GTM workflow for one real company — an enrichment waterfall, signal-based outbound, an inbound router, or a CRM cleanup — with a 30-minute call each week. It also notes that Zach Vidibor will teach in the upcoming Cohort 5.
Waxman spent an hour with Swan AI founder Amos Bar-Joseph testing the company's claim that it went from zero to 200 customers with three founders and no sales, customer success, or support staff.
Read more
The post frames the conversation as an interview about where a company's context layer should live, with the full hour posted to YouTube.
In their newsletter roundup, Waxman reports that Sabrina Carver reached the Clay Cup semifinals and describes the competition format: 25 minutes to clean a CRM, find the ICP, score the accounts, and draft outbound, with no prebuilt workflows and no LLM.
Waxman describes the context-layer setup at WorkSpan, where the marketing organization standardized on Claude plus a context layer, and relays Sam Gong's split between a brand layer anyone in marketing can write to, a governed canonical layer for product and GTM strategy with a release cycle, and deterministic systems of record answered through a semantic layer.
Waxman leads a newsletter issue with OpenAI's Astra model, describing a test in which Astra was handed 15 customer emails, could not find the company's reply standard, and sent nothing, while the older model sent all 15.
Read more
They argue that an agent's ceiling now depends on whether the rule it searches for was ever written down.
Jared Waxman writes for people putting AI to work in sales and marketing.
Read more
The quickest way into the work is the weekly roundup, GTM Engineer Pulse, which mixes their interviews with reader-submitted items about what teams are actually shipping.
Longer reads feature hour-long conversations with operators.
The strongest thread is that models and agents are not the advantage: where a company’s context lives, who owns it, and whether anyone wrote down the rule an agent needs decide the outcome.
To learn the practice, join a cohort of GTM Engineer School, where each participant builds one real workflow for one real company.
Spencer analyses how AI is changing B2B go-to-market, landing on the theme Waxman returns to: the flashy part of the stack is not what decides whether anything works.
Read more
Spencer argues that an AI agency lives or dies on delivery, not the funnel, and makes the case that Salesforce turned its platform into something agents and coding tools drive directly. Like Waxman’s work on context layers, Spencer looks beneath the tooling to the infrastructure and the work.
Spencer writes from inside an AI agency while managing operations at an appliance repair shop, framing his posts around market moves and the business of running an agency, including a stock-and-platform read on Salesforce.
Waxman’s material takes a more instructional approach: recorded operator interviews, a weekly roundup, and a cohort course where people build one workflow for one real company.
Go to Spencer for the agency and market view, and Waxman for how to structure the context and rules an agent needs.
Ruha publishes concrete GTM AI systems rather than commentary: signal-based research feeds, a Claude Code skill that checks accounts against a written rule, and a long set of go-to-market rules broken out by ARR stage.
Read more
She and Waxman both treat AI work as workflow design, insisting that input quality and the written process determine the result.
Ruha organises her advice into staged rule lists and lead-generation builds, with detailed material often behind a video or agent download.
Waxman focuses on context layers, written standards, and where a person still approves an action, and offers the learning through a free cohort.
Read Ruha for a step-by-step lead-gen system at a particular revenue stage, and Waxman for arguments about where company context should live.
McNamara writes about the same shift Waxman describes from the RevOps and GTM ops side: technical work is moving toward describing a process precisely enough for an agent to build it, with failure modes and guardrails—not implementation—the hard part.
Read more
Waxman makes the parallel point that an agent’s ceiling depends on whether anyone ever wrote down the rule it needs.
McNamara takes an organisational, career-facing angle.
She tracks how AI leadership roles are appearing inside RevOps and what those mandates look like, and walks through her own Claude Code and Clay build.
Waxman focuses on where context lives and where humans stay in the loop, delivering that work through interviews, a roundup, and a school.
Choose McNamara to see how roles and reporting lines are changing, and Waxman for decisions about context layers and agent approval.
Placement decision: use Jared Waxman for a practitioner audience already working on GTM engineering, AI context layers, and agent workflows—not for a broad SaaS leadership list.
Read more
The only measured audience figure in the posts is 8,873 LinkedIn followers.
The content points toward an operator audience: the Swan AI post asks readers what stack they use to build context for AI; the Astra post asks what rule an agent looked for and could not find; the GTM Engineer Pulse post asks which meeting listeners would decline now that they would have taken 18 months ago; and the GTM Engineer School post asks where teams are building their context layer.
Those questions target people already close to GTM systems and AI-assisted sales and marketing work.
For a talk or advisory session on context layers and GTM engineering, Jared Waxman is a fit if the buyer can bring the qualifying audience or review the host's audience list.
If the campaign depends on seniority, industry, or company size, ask whether they have audience survey results, follower demographics, or recent registration lists from comparable sessions.
LinkedIn followers
8,873
LinkedIn profile audience count captured in the account record for https://www.linkedin.com/in/jwaxman · 2026-10-08
Jared Waxman's four posts cover editorial and program-related topics: weekly GTM engineering roundups, a recorded conversation with a vendor founder, and a note about GTM Engineer School.
Read more
They don't advertise a rate card, sponsored-placement format, branded-series offer, or inquiry route for paid partnerships.
One adjacent program is Cohort 5, described as free and "not consulting, no promised deliverable," with buyers supplying a concrete problem and joining a 30-minute weekly call; entry is by commenting CAPSTONE.
It's a participation program, not a sponsorship format, and its terms come from the creator rather than a negotiated paid engagement.
Brands should ask directly about partnership availability, format, and terms; the posts provide no published route or menu.
Strong fit, provided the buyer settles two conditions first.
Read more
Brief
A B2B SaaS company building its GTM AI context layer and agent workflows wants a talk or advisory session for the revenue marketing, RevOps and sales operations people who will run the pilot.
Reasons
The session topic is already at the core of the published material. One post explains what a context layer contains: a brand layer anyone in marketing can write to, a governed canonical layer for product and GTM strategy with a release cycle, and a deterministic layer of systems of record answered through a semantic layer. Jared Waxman's observation is that most companies say "context layer" and mean one of the three without knowing which. That gives clients a diagnostic they can apply in the room.
The adoption question the client will face has a published answer to test. The same post separates builders from consumers and reports that at WorkSpan the campaigns team and RevOps went in first, while everyone else got a chat interface with the plugin installed and never saw the plumbing. A client choosing its first two functions gets a starting point rather than a blank page.
The scoping rule is directly usable. The advice quoted from Zach Vidibor is to avoid making the project "build a company brain" and instead pick one problem with a deadline. The example is WorkSpan loading its sales kickoff because reps had to position a feature to a services partner one day and a cloud partner the next. That gives a working session a short, fundable pilot shape.
The agent-approval question comes with two concrete patterns. One company's agents earn autonomy through their acceptance rate; another keeps a person on every send. Joe Rhew's argument, as reported, is that the gate gets worse as the system improves. A client designing approvals can use that as a real choice rather than defaulting to a tool's settings.
The documentation point suggests a small first deliverable. In the account, an OpenAI model given 15 customer emails found no reply standard and sent nothing. That turns an abstract governance worry into an audit a client can approve before any build.
The operator niche suits a session where the client brings its own team. The posts' questions target people already close to GTM systems and AI-assisted sales and marketing work.
Conditions
Jared Waxman's posts describe no paid talk or advisory format. The only adjacent program, the GTM Engineer School cohort, is free, "not consulting, no promised deliverable," and entered by commenting on a post. An advisory session would be a new arrangement, not a purchase from a published menu; discuss terms directly.
The only audience figure is 8,873 LinkedIn followers; there is no survey, registration list or demographic detail showing seniority, company size or function. If the session depends on a particular seniority mix, ask the host for audience details or plan to fill the room from the client's own list.
When to choose a specialist
Poor fit based on the posts reviewed.
Read more
Brief
A CRM or cloud vendor wants a broad sponsored content series to reach a general B2B software buying audience of tens of thousands, purchased through a published sponsorship rate card with disclosed placements.
Reasons
The measured audience is 8,873 followers on one channel. The posts don't specify the audience's roles, seniority or company-size mix, so ask for that breakdown before committing to a campaign aimed at a large general buying audience.
The material reviewed includes no rate card, sponsored-format description, disclosed sponsor example, branded-series offer or partnership inquiry route. Ask whether a sponsorship product and placements are available before planning the activation.
The recurring format is curation, not exclusive placement. The weekly roundup mixes original items with other people's work and news, naming vendors and other practitioners side by side. A sponsor added to that list competes with every other item for attention, which works against a short, high-visibility branded placement.
The longer formats send the audience elsewhere: the hour-long conversations point to full recordings on YouTube. That suits people choosing to spend an hour on a practitioner problem and a niche audience, but not the skim-and-leave exposure a broad awareness buy usually wants.
The clearest working program in the material is explicitly not a commercial engagement. Cohort 5 is free, runs on one 30-minute call a week, promises no deliverable and admits participants through a public comment. Presenting it as a sponsorship product would misrepresent it.
A worked campaign example
Angle
Run the session as a diagnostic the client performs on itself, using the classification and scoping questions Jared Waxman already puts to this audience. Keep the promise narrow and checkable: by the end, the client has named which context layer it is building, who owns it and one problem with a deadline that the first pilot will solve.
Brief
The good-fit brief: a B2B SaaS company building a GTM AI context layer wants a session that ends with a decision, not a vendor tour.
Deliverable
A one-page layer map naming each of the three layers and its owner, plus a single scoped pilot with a deadline and a named approver for agent actions.
Format
A 60-minute recorded conversation for the client's revenue team, following the shape of the hour-long sessions already published, with a working block drawn from the cohort format: one real problem, one short call a week and a working artefact at the end.
Session flow
Open with the question from the Swan AI conversation: where does your context live today, and who can read it? The stated failure modes make the question useful: a code repository serves agents but not people; a docs tool serves people but not agents.
Classify the client's existing material into the three layers and name an owner for each. Check whether the governed product and GTM strategy layer has a release cycle; the material says most companies leave that layer undefined.
Split builders from consumers and pick the two functions that go in first. The published example is campaigns and RevOps, with everyone else receiving a chat interface and never seeing the plumbing.
Scope one problem with a deadline. In the source's words, do not make the project "build a company brain." Use the sales-kickoff example: reps coming out of an event had to position one feature to two different kinds of partner.
Write down a rule the agent will look for. Use the fifteen-email example to capture a reply or positioning standard that currently lives in someone's head.
Set the approval gate: choose whether agents earn autonomy through their acceptance rate or a person approves every send. Name who reviews the gate as accuracy changes.
Close with the audience question from the GTM Engineer School post: where is your team building its context layer today?
Success measures
Name the single context layer the client is building first and who owns it.
Agree on one problem with a deadline, write it down and list the functions involved in the first wave.
Identify at least one unwritten reply or positioning rule to document.
Record a named approver and a review point for agent autonomy.
Produce a client-owned artefact, not a recording the client has to reinterpret.
Watch outs
Treat the Swan AI headcount claim and the vendor context-layer claim as vendor claims relayed in a post, not verified results. Ask for the client's own numbers before reusing either claim in the room.
Present the three-layer taxonomy as one marketing organisation's setup, not an industry standard. Treat it as one workable model and expect the client's version to differ.
If the recording goes public, agree where it will live and who holds the rights; the published conversations point to YouTube recordings.
Keep the two functions small in the first wave. The material warns against naming half the company as builders.
Use this to decide fit for a talk or advisory session. Whole-page acceptance, independent review and release are separate steps.
Why this angle
Each step tests a client decision against material Jared Waxman has published, rather than simply restating the topic. Close with the audience question used in the posts to match how they already speak to this group.
Choose Sara when the buyer's main challenge is deciding who should own GTM AI adoption and how it fits into RevOps.
Read more
Her post compares two emerging models: an AI-native RevOps leader who owns the operating model and builder teams, or a separate GTM AI leader alongside RevOps.
She also describes the shift toward writing precise process specifications that agents can build from, including the need to define exceptions, guardrails, and tests.
That makes her a closer fit for an organizational-design session or a working session to turn an existing process into a buildable specification.
The posts focus on RevOps leadership and process design, not Jared Waxman's three-layer context model or their advice to start with a small, deadline-bound pilot. Buyers seeking that framing and a concrete first-use-case discussion may prefer Jared.
Choose Brigitta when the buyer's immediate priority is a practical AI-assisted lead-generation workflow.
Read more
Her example lays out a sequence for selecting buying signals, checking accounts against a two-signal threshold, showing evidence for each qualification, and shaping outreach from real messages. That offers a direct starting point for a demand-generation team seeking a campaign or workflow session.
The posts center on lead generation and scaling GTM rules by revenue stage. They don't address context-layer ownership, governed company knowledge, or introducing agent workflows across builders and everyday users—central to Jared's fit for campaigns focused on those needs.
Choose Shane if the buyer wants a critical discussion of AI-agency delivery or the implications of Salesforce's shift toward APIs, agent tools, and Slack-based workflows.
Read more
His examples take a strong operator's point of view: one argues that service delivery matters more than agency funnels, while another describes building agents and skills around Slack and business systems. That could suit a session focused on delivery discipline or a specific Salesforce-and-Slack platform direction.
These posts don't offer Jared's explicit framework for classifying context layers, choosing initial builder and consumer groups, and scoping one pilot problem. They're also less direct as a guide for a general B2B SaaS team deciding how to structure its GTM AI context work.
Jared Waxman's posts here include four LinkedIn posts and an account with 8,873 followers as of 8 October 2026.
The posts cover an hour-long conversation with Swan AI founder Amos Bar-Joseph about where a company's context layer lives, a roundup issue led by an OpenAI model test and the documentation question behind it, an interview-based note on the GTM Engineer School setup, and a further GTM Engineer Pulse edition on agent approval, CRM agent releases and cohort logistics.
The posts carry no publication dates; they reference a cohort starting 26 October and a January 2025 baseline for one figure.
Two posts summarise hour-long recorded conversations and link to YouTube recordings; guest views are reported as they appear in the posts.
Roundup items naming other practitioners and companies are summarised in the posts.
The similar-creator comparisons in the discovery section refer to posts from Shane Spencer, Brigitta Ruha and Sara McNamara; their internal page links are destinations rather than source material here.
No transcript, audience survey, registration list, rate information or partnership inquiry route is published in the posts here.
The posts here do not mention a book, show, sponsorship, affiliate arrangement or paid client engagement, and make no claim of one.