Pasha Irshad works on HubSpot-focused RevOps and go-to-market engineering, including custom automation and practical CRM implementations. Their posts explain how to solve operational problems, test AI agents, and use HubSpot data in ways that support marketing and sales teams.
Most useful for concrete HubSpot automation and AI implementation explanations, including limits, testing, and tool selection.
Best for
HubSpot operations, CRM automation, or applied AI campaigns needing practical implementation detail.
Not for
Paid creator sponsorships for broad audiences needing rate cards or demographic guarantees.
Key takeaways
An IT-posture agent first misclassified a private equity firm, treating technology leadership roles as internal IT.
They used a tested Python process to write company-to-event associations through HubSpot's API.
They reserve paid AI agents for contextual interpretation, not deterministic rules.
Check first: Ask for a build walkthrough or working files before treating posts as proof of delivered code.
Main ideas
Define the output before deciding whether AI should do the work · Test agents on real accounts and expect to repair their reasoning · Context maintenance is the precondition for CRM AI
Pasha Irshad works on HubSpot-focused RevOps and go-to-market engineering.
The posts show a practical builder’s perspective: identify an operational problem, choose an appropriate tool, test the result, and explain what went wrong or what it costs.
Examples include an agent that classifies companies’ IT posture, a Python process that connects event attendees’ companies to HubSpot event records, and a Slack alert for high-intent leads that need attention.
Follow them for concrete explanations of HubSpot automation and AI, including the limits and testing considerations—not just product announcements.
Irshad makes implementation decisions legible.
In the IT-posture agent example, the first test misclassified a private equity firm because the agent treated technology leadership roles as an internal IT department.
They explain the rule changes that addressed the error and the rerun, rather than presenting the build as successful from the start.
Other posts show the same practical judgment: use an API when workflows cannot create the required event association, and reserve paid AI agents for contextual interpretation rather than deterministic rules.
Specific build details, tested corrections, and clear tool-selection reasoning help readers understand what the automation can do and where it needs care.
Pasha Irshad's ideas follow a simple sequence: define the outcome, supply the context, then choose the smallest tool that can deliver it.
Read more
Across these posts, they argue that HubSpot teams overpay for AI when a task has a repeatable answer, agents need explicit rules and real-account testing before running at scale, and CRM automation works only when someone keeps business context current. The posts are mostly worked examples, so the arguments emerge through build decisions, failures, and repairs rather than general commentary.
Define the output before deciding whether AI should do the work
Pasha draws a line between deterministic and probabilistic work.
Read more
Deterministic tasks have a rule and a repeatable answer, so they belong in workflows, calculated properties, custom code, or the API, where the same result is produced every time without spending credits. Probabilistic tasks require interpretation, and that is where a HubSpot Data Agent can earn its cost: summarizing account research, reading buyer intent, identifying likely pain points, recommending a sales angle from a transcript, or turning unstructured CRM activity into something usable.
Setting an account tier when employee count exceeds 500 and annual revenue exceeds $50M is deterministic; it should not run on a paid agent.
Reading industry, headcount, website, intent signals, CRM notes, and recent activity to summarize why a company may need managed IT, cybersecurity, cloud, or help desk support is probabilistic.
A 40-person law firm and a 700-person manufacturer researching cybersecurity are in different situations, and an agent can capture the surrounding context in a property.
Test agents on real accounts and expect to repair their reasoning
Pasha treats an agent build as a test-and-fix loop, not a one-time prompt.
Read more
An early version can misread job titles, company types, or operating models.
The fix is to write explicit inclusion and exclusion rules, add entity-type checks that adapt the agent's reasoning to each company type, and rerun the same accounts.
They also warn that anyone putting an agent live needs to understand the credit spend first.
A custom agent in HubSpot's Agent Builder was designed to classify a target company's IT posture for a managed IT services client.
Its first test on a private equity firm returned “Internal IT Team, estimated headcount of 10”—wrongly.
The firm has a Center of Excellence with partners and VPs of Technology & Innovation and Data & AI.
The agent saw technology titles and interpreted them as an internal IT department.
The fix was to count only operational IT roles, exclude data science, AI, analytics, and technology titles within a Center of Excellence or an investment firm's operating bench, and identify private equity and venture capital firms first.
Rerunning the same company produced the expected result.
Context maintenance is the precondition for CRM AI
Pasha's product commentary returns to a shift in who does the work: people define context and state intent, while HubSpot captures signals and updates the CRM.
Read more
The argument is that AI cannot solve a team's problems without context. They assess product releases by whether they reduce the steps end users must take to keep business context current.
After HubSpot's Unbound 2026 keynote, Pasha highlighted a Smart CRM that captures and syncs context from calls, emails, and meetings rather than relying on representatives to maintain records; a Context Home that surfaces gaps in business context; Breeze's shift from AI assistant toward an orchestration layer, where users describe an outcome and specialized agents handle parts of the work; and Marketing Studio updates that let a weak segment or unanswered leads become the starting point for a campaign. A separate post on the Conversational Enrichment public beta applies the same test to a feature that enriches contact and sales data from unstructured email conversations, noting which plans include which signals and how beta enrichment events interact with credits.
The Unbound post states the context-first position directly. The enrichment post applies the same test to a specific beta release and separates out the credit caveats.
When the platform path doesn't fit, build the API route and verify it
Pasha treats HubSpot limitations as design constraints rather than dead ends, preferring the lightest tool that works on software the team already owns.
Read more
The method: find the unsupported path, build it, test it against real records, and quantify the result so it can be reused. The same preference is evident in choosing a script and webhook over a new dashboard tool.
A client wanted a direct Company-to-Marketing Event association written back to the CRM when a contact attended an event.
A segment could identify the companies, but it produced only a filtered audience, not the association.
Manual association becomes impractical beyond about 25 companies; the client lacked Data Hub, so custom-coded automation was unavailable, and available triggers limited other automation options.
Pasha found that HubSpot's API supports the association directly, using type 1387 for Company to Marketing Event and 1386 in reverse, then built and tested a Python process.
The test resolved 16 contacts to 10 unique companies, created eight new associations, preserved two existing ones, and verified all 10 without duplicates.
In a separate post, a daily Slack alert for high-intent leads uses HubSpot's agent CLI and a Slack webhook, with no new tool to buy or integration to maintain.
Overcomplicated, overfunded stacks pushed the work toward RevOps
Pasha traces the origin of the current work to ten years running demand generation as a VP of digital and a recurring observation: the smartest marketers were often constrained by their technology stacks.
Read more
Those stacks were overcomplicated, under-resourced, and most of the time overfunded.
Working on those problems led them into RevOps and deeper into HubSpot.
That background helps explain why the published work reads as build-and-test case studies rather than product commentary.
The post does not name a client or a specific result. It works as context for the implementation examples elsewhere and closes by inviting anyone dealing with a similar stack problem to reach out.
The clearest example here of a custom HubSpot agent build in practice, including what many write-ups leave out: the first version got a company wrong, and the fix is shown.
Read more
The client is a managed IT services provider.
The agent's job was to research any target company and classify its IT posture: whether IT is run internally, whether there is an incumbent managed service provider, and whether a renewal window makes outreach worth timing.
That research had been manual — LinkedIn, careers pages, company sites, news, and sometimes IRS Form 990 filings for nonprofits — at 15 to 20 minutes per account.
The first test was a private equity firm, and the agent answered "Internal IT Team, estimated headcount of 10," which was wrong: the firm has a Center of Excellence with Partners and VPs of Technology & Innovation and Data & AI.
The agent saw technology titles and read them as an internal IT department, missing that those roles support portfolio companies rather than the firm's own help desk and infrastructure.
The fix was an explicit rule counting only operational IT roles (IT Directors, Systems Administrators, Infrastructure Engineers, Help Desk) and excluding data science, AI, analytics, and technology titles inside a Center of Excellence or an investment firm's operating bench, plus an entity-type check that identifies private equity and venture capital firms first and adjusts the reasoning.
Rerunning the same company produced the expected result.
The agent now runs across a full Tier 1A account list and sorts companies into internal team, outsourced with a confirmed vendor, outsourced with an unknown vendor, or unclear; for nonprofits it pulls the IRS Form 990 and surfaces the actual IT vendor and annual spend.
The post also warns that anyone pushing this live has to test it and understand the credit spend first.
For brands planning a campaign
For a brief on custom automation, this is the strongest starting point: a scoped build, a specific failure, a defined repair, rerun results, and a case for reuse across many accounts. A campaign could ask for a build walkthrough covering setup, credit allocation, and testing, or use the PE-firm misclassification as a discussion topic for a technical audience.
Answers a narrow, practical question — how to create a CRM relationship HubSpot will not build for you — and backs it with measured test results.
Read more
A client wanted a simple outcome: when a contact attends a marketing event, identify their associated company and connect it to the same event automatically.
Segments could identify companies through associated contacts' attendance, but that only produces a filtered audience, not the association.
The client wanted the relationship written back to the CRM as a direct Company-to-Marketing-Event association visible on the event record.
Manual work breaks down past about 25 companies; the client lacked Data Hub, ruling out a custom coded workflow, and workflows struggled with limits on available triggers.
The API supported the requirement directly, with association type 1387 for Company to Marketing Event and 1386 for the reverse.
The built and tested Python process takes the contacts on a marketing event, confirms the qualifying participation state, finds each contact's primary or only company, checks whether the company is already connected, creates any missing association through the API, then verifies and logs the result.
In the test, 16 contacts resolved to 10 unique companies, creating eight new company-to-event associations, preserving two existing ones, and verifying all 10 with no duplicates or manual mapping.
The client is reusing it across multiple events.
For brands planning a campaign
A tested, repeatable deliverable reused beyond a one-off fix — a good fit for a campaign about automation that lasts beyond the first event. The API-versus-workflow reasoning could also anchor a technical explainer or client-facing decision aid.
Offers a decision rule for what belongs in workflows and what merits a paid HubSpot agent — usually the first question when a client asks to add one.
Read more
Start by defining the output, then choose the tool.
Deterministic work has a clear rule and repeatable answer — for example, "if employee count is over 500 and annual revenue is over $50M, set Account Tier to Tier 1" — and belongs in workflows, calculated properties, custom code, or the API.
These produce the same output every time and do not incur a per-run cost.
Probabilistic work requires interpretation, such as reviewing a company's industry, employee count, website, buyer intent signals, CRM notes, and recent activity to summarize why it may need managed IT, cybersecurity, cloud, or help desk support.
That is where the Data Agent earns its credits: summarizing account research, interpreting buyer intent signals, identifying likely pain points, recommending a sales angle from a transcript, and turning unstructured CRM activity into a usable insight.
A 40-person law firm researching cybersecurity is in a different situation from a 700-person manufacturer researching the same topic, the post notes, and it shows the kind of property the agent can write to capture that context.
It recommends controlled tests on a small set of records to check that the output matches the intent and the monthly credit budget does not disappear in one run.
For brands planning a campaign
Shows how Pasha Irshad scopes AI work and controls cost before writing anything, a useful differentiator for a custom automation brief. A campaign could ask for a deterministic-versus-probabilistic review of the client's own use cases as a concrete deliverable.
A short, finished solution for anyone tired of dashboards nobody opens, and a clear look at how Pasha chooses the lightest tool for the job.
Read more
The post opens with the "graveyard of dashboards" — random reports fused into a dashboard that has not had a view in a year — and a question often skipped in AI discussions: is this useful, and what use case does it serve?
One frequent request is visibility into the sales pipeline and rep activity, especially high-intent leads.
After reviewing HubSpot's agentic options, Pasha built an alerting system using HubSpot's agent CLI and a Slack webhook.
Every morning, a short script checks which high-intent leads (demo requests, contact-sales form submissions) are still at "Open" or have no owner, then posts to a Slack channel with the contact name, form submission, lead status, contact owner, and a synopsis of what has happened.
Time of submission was added later.
The post includes an anonymized screenshot of the returned data.
The system helps spot gaps, anomalies, and automation problems without logging into a dashboard.
It is lightweight: no new tool to buy, no integration to maintain, just a script, a scheduler, and a webhook URL.
The post asks readers what they would want in a daily alert.
For brands planning a campaign
A compact demonstration of operational thinking and a low-cost build, well suited to a walkthrough or build-along. The closing question invites comments or direct replies.
Shows how Pasha evaluates a new HubSpot feature, detailing each feature and the billing details that often surface later — useful reading before a buyer turns something on.
Read more
The post covers Conversational Enrichment for Email, launching in public beta out of Unbound 2026, an AI feature that captures and enriches contact and sales data from unstructured email conversations.
It separates out the practical details: contact records stay current automatically, and Contact Enrichment from Email is available across all Breeze Intelligence plans, updating CRM details based on inbound email conversations; Returning to Office Detection and New Contact Extraction are included with the 1,000- and 10,000-credit Breeze Intelligence plans; and beta enrichment events do not currently consume credits directly, though additional enrichment applied to those records based on Breeze settings may still consume credits.
The post argues that without context, the agentic future for HubSpot and other CRMs is much harder to realize, and that this feature gives end users one less step to keep context up to date.
For brands planning a campaign
Useful for campaigns aimed at teams weighing new HubSpot AI features: the credit caveats are specific and checkable. It also pairs well with the deterministic-versus-probabilistic post as a single point of view on adding AI to an existing instance.
A compact statement of Pasha's view of how HubSpot is changing and where people still matter, useful orientation before the individual builds.
Read more
Written after the keynote and product spotlight, the post distills the shift into two lines: previously, humans operated HubSpot and HubSpot stored data; increasingly, humans define context and state intent, HubSpot captures signals and performs work, and the CRM updates itself.
Its standouts: Smart CRM is becoming more autonomous, capturing and syncing context from calls, emails, and meetings rather than relying on reps to maintain records; a new Context Home surfaces where business context is incomplete; Breeze is moving beyond an AI assistant toward an orchestration layer, where users describe the outcome they want and specialized agents handle parts of the work across marketing, sales, and service; and Marketing Studio updates tighten its connection to CRM context, performance signals, and the agent layer.
As a result, an underperforming segment, weak AEO visibility, or leads that are not being followed up on can increasingly become the starting point for a campaign.
The post closes on the keynote's emphasis: without context, AI will not be the answer to all your problems.
For brands planning a campaign
A ready-made point of view for a campaign angle, bridging conference news and the hands-on implementation posts elsewhere in this section. A good fit for a campaign seeking a commentator on HubSpot's direction, not only a builder.
In their own words, why they work on HubSpot implementation.
Read more
The post traces a decade running demand generation as a VP of digital, where the smartest marketers were often constrained by tech stacks that were overcomplicated, under-resourced and, most of the time, overfunded.
Seeing those problems day in and day out led them into RevOps and deeper into HubSpot, prompting them to share client results like those shown elsewhere in this section.
They invite anyone facing similar challenges to reach out.
For brands planning a campaign
Background and positioning material rather than a technical demonstration. For case-study planning, ask for the client name and the specific result behind the post's claims.
Building, testing, and repairing HubSpot AI agents
Pasha Irshad describes building a custom agent inside HubSpot's Agent Builder for a managed IT services client.
Read more
The agent researched target companies and classified their IT posture: an internal IT team, an incumbent managed service provider, or a renewal window worth timing outreach around.
It replaced a manual process that took 15 to 20 minutes per account and drew on LinkedIn, careers pages, company sites, news, and sometimes IRS Form 990 filings for nonprofits.
The first test, on a private equity firm, came back wrong: the agent saw technology titles and classified them as an internal IT department of about 10 people, missing that those roles supported portfolio companies.
The fix was an explicit rule to count only operational IT roles (IT directors, systems administrators, infrastructure engineers, help desk) and exclude data science, AI, analytics, and technology titles inside a center of excellence or investment-firm operating bench.
An entity-type check also identifies private equity and venture capital firms first.
Test an agent on a small set of accounts before launch, write explicit inclusion and exclusion rules, and add an entity-type check so its reasoning changes by company type. Pasha also recommends estimating credit spend before running an agent across a full account list.
For brands planning a campaign
For a brief calling for custom automation expertise, this shows the work of building, breaking, and repairing an agent inside HubSpot's own tooling—not just advising on it. A brand selling agent or AI enablement could use the misclassified private equity firm as a worked example of why pilot design and clear rules matter, and the corrected run as evidence of a test-and-fix loop.
Pasha Irshad frames HubSpot Data Agent use cases by first defining the output.
Read more
Deterministic work—such as setting an account tier when employee count and revenue cross fixed thresholds—belongs in workflows, calculated properties, custom code, or the API, because the answer is repeatable and should not cost credits.
Probabilistic work—such as reading industry, headcount, website, buyer intent signals, CRM notes, and recent activity to summarize why a company might need managed IT, cybersecurity, cloud, or help desk support—is where the Data Agent earns its cost.
Examples include summarizing account research, interpreting buyer intent, identifying likely pain points, recommending a sales angle from a transcript, and turning unstructured CRM activity into a usable insight.
One illustration: a 40-person law firm and a 700-person manufacturer are both researching cybersecurity, but the same intent signal means different things to each.
The agent can write a property that captures the surrounding context.
For brands planning a campaign
This positions Pasha Irshad as someone who scopes AI projects by outcome rather than tool, a fit for a brand brief on automation architecture and credit budgeting. It also cautions against framing every CRM task as an AI task: a campaign message focused solely on autonomy would sit awkwardly with this approach.
A client wanted a direct Company-to-Marketing Event association written back to the CRM and visible on the Marketing Event record when a contact attended an event.
Read more
A segment could identify companies through their contacts' attendance, but that produced only a filtered audience, not the association.
Manual association breaks down when more than 25 companies attend an event, and the client had no Data Hub, so custom-coded automation was unavailable.
Automations were also limited by their available triggers.
Pasha Irshad found that HubSpot's API supports the association directly (association type 1387 for Company to Marketing Event and 1386 in reverse) and built and tested a Python process: take the contacts on the event, confirm qualifying participation, find each contact's primary or only company, check whether the company is already connected, create any missing association through the API, then verify and log the result.
In the test, 16 contacts resolved to 10 unique companies, creating eight new associations, preserving two existing ones, and verifying all 10 without duplicates.
Resolve contacts to companies, check for existing associations before writing, create only what's missing, then verify and log every result so the process can be rerun across events.
The client lacked Data Hub, ruling out custom-coded automation; the API was a workaround for a platform limit, not a preference.
For brands planning a campaign
This is a useful sample for a brief involving CRM data hygiene or marketing operations.
It shows Pasha Irshad reading platform constraints, finding an unsupported path, and testing it against a real record set.
They also quantify the result (16 contacts, 10 companies, eight new associations) rather than describing the build alone.
Pasha Irshad describes a graveyard of dashboards that have not been viewed in a year and recurring requests for visibility into sales pipeline and representative activity on high-intent leads.
Read more
After reviewing HubSpot's agentic options, they chose to build an alerting system using HubSpot's agent CLI and a Slack webhook.
Every morning, a short script checks which high-intent leads, such as demo requests or contact-sales form submissions, are still sitting at Open or have no owner.
It posts the contact name, form submission, lead status, owner, and a synopsis of what has happened to a Slack channel.
Submission time was added later.
The system spots gaps, anomalies, and automation failures without requiring anyone to log into a dashboard, and requires no new tool to buy or integration to maintain.
The post asks the audience what they would want in a daily set of alerts.
For brands planning a campaign
For a brief about internal enablement, RevOps visibility, or go-to-market engineering, the appeal is a low-overhead build using tools the team already owns. Pasha invites the audience to suggest alert content, a prompt a brand could echo in a co-created campaign format.
Reacting to HubSpot's Unbound keynote and product spotlight, Pasha Irshad describes a shift from humans operating HubSpot while it stores data to humans defining context and stating intent while HubSpot captures signals, performs work, and updates the CRM itself.
Read more
They highlight a Smart CRM that captures and syncs context from calls, emails, and meetings rather than relying on representatives to maintain records; a Context Home that surfaces gaps in business context; Breeze's shift from AI assistant toward an orchestration layer, where users describe the outcome and specialized agents handle parts of the work; and Marketing Studio updates that tie campaigns more closely to CRM context, performance signals, and the agent layer, so a weak segment or unanswered leads can become the starting point for a campaign.
The through-line is that without context, AI will not be the answer.
A separate post covers the Conversational Enrichment public beta, which enriches contact and sales data from unstructured email conversations across Breeze Intelligence plans, with returning-to-office detection and new contact extraction on higher-credit plans.
Beta enrichment events do not consume credits directly.
For brands planning a campaign
A brand whose brief rests on AI messaging should note that Pasha Irshad's public framing puts context ahead of autonomy. That makes this a credible angle for campaigns about data quality, CRM hygiene, and enrichment, and a less natural fit for messaging that treats agents as a substitute for maintaining customer context.
A demand-generation background that moved into RevOps
Pasha Irshad traces the current work to ten years running demand generation as a VP of digital and a recurring observation: the smartest marketers were constrained by technology stacks that were overcomplicated, under-resourced, and most of the time overfunded.
Read more
Working on those problems led into RevOps and deeper into HubSpot.
For brands planning a campaign
This is the stated path behind the implementation posts: a demand-generation operator who moved into operations, which helps explain why the published work reads as build-and-test case studies rather than product commentary.
For a brand assessing expertise for a custom automation brief, it is the clearest statement of the practitioner's background.
It also supports campaign angles on stack sprawl and consolidation, the problem Pasha Irshad says drove the shift.
After attending a series of HubSpot Inbound and Unbound conferences, Pasha Irshad co-hosted a happy hour and ping pong event during Unbound with Kamil Rextin and 42 Agency.
Read more
They described the arrangement as trading a few hundred hours of go-to-market engineering and HubSpot work for the co-hosting slot. The event theme was stated as brand being hot.
For brands planning a campaign
This is a concrete signal of in-person presence at HubSpot's flagship event and of working alongside an agency partner on a side event. A brand considering event-adjacent sponsorship or a partner-led activation has a recent example of Pasha Irshad organizing and promoting one.
The posts describe one named brand partnership and consistent public work with HubSpot.
Read more
Pasha Irshad co-hosted a happy hour and ping pong event during HubSpot's Unbound conference with Kamil Rextin and 42 Agency, exchanging a few hundred hours of GTM engineering and HubSpot work for the co-hosting slot.
HubSpot features in most of the work described — custom agents in Agent Builder, the Data Agent, the Marketing Events object, and product launches.
Ask whether this HubSpot work is sponsored, affiliate, or employment-related.
The posts also describe client work for an unnamed managed IT services company; ask for the client's name.
HubSpot
Mention
HubSpot is the platform behind most of the work described: a custom agent built in Agent Builder that classifies IT posture for a managed IT services client, guidance on which Data Agent tasks justify HubSpot credits, a Python process built on HubSpot's association API for a client without Data Hub, and a morning alerts script using HubSpot's agent CLI plus a Slack webhook.
Read more
The posts also cover HubSpot's Unbound keynote and the Conversational Enrichment beta.
Pasha Irshad co-hosted a happy hour and ping pong event during Unbound with Kamil Rextin and 42 Agency.
Read more
They promised a few hundred hours of GTM engineering and HubSpot work to Kamil Rextin and 42 Agency in exchange for the co-hosting slot, with the event theme "BRAND IS HOT."
Pasha Irshad’s posts read best as a build log for HubSpot teams.
Read more
Start with the custom agent built in HubSpot’s Agent Builder for a managed IT services client: it shows the whole loop, from the job to be done and the first test that came back wrong to the rule that fixed it and the corrected rerun.
The Data Agent post explains how they decide whether a task belongs to an agent at all; the Marketing Events post applies the same instinct to a platform gap, with an API workaround built, tested, and counted against real records.
Their Unbound commentary and daily Slack alert round out the picture: keep business context current, then automate the smallest useful piece.
The tags below point to this material by subject, specialty, format, and assumed audience.
Three adjacent writers are listed at the end; each overlaps on HubSpot and revenue operations but takes a different approach.
Writes about the same HubSpot and revenue-operations territory, making much the same point about credits: the useful question is what the work returns, not how few credits it burns.
Read more
Both explain HubSpot features through the cost and readiness questions teams face before switching them on, and examine releases such as Revenue Hub in light of the data underneath.
Etienne Turner weighs costs and prerequisites before making a recommendation, using concrete per-action credit figures and a rollout sequence.
Pasha takes a builder’s approach to the same questions: an agent built inside HubSpot, a rule added after a wrong result, and an API association process tested and counted against real records.
Readers weighing an AI or Revenue Hub decision will find the strategic framing in one and the implementation detail in the other.
Both write about HubSpot’s product direction and the Unbound announcements, including Marketing Studio, a self-updating CRM, and Breeze handling more of the work.
Read more
Both argue that context, not tooling alone, makes AI useful.
Yamini Rangan writes from the standpoint of what HubSpot built, describing the launch from the stage and using it to explain where the product is going.
Pasha writes from the customer side, assessing the same release against the work a team must do first and testing specific features such as contact enrichment against plan limits and credit rules.
One offers the announcement and intent; the other, an implementer’s read.
Both make the case that agents sit on top of the data layer, so enrichment, ownership, identity resolution, and routing must be sound before AI can do its work.
Read more
They describe what goes wrong when records are half-filled or duplicated, and point to the unglamorous fixes rather than the agent itself.
Corrina Owens makes the case in the abstract and through a vendor platform, laying out four diagnostic questions and the order of operations a revenue team should follow.
Pasha tackles the same question with a shipped build: a party, a test failure, a rewritten rule, and a verified result across a live account list.
Read Corrina Owens for the pre-rollout checklist and Pasha for what it looks like in practice.
Pasha Irshad’s LinkedIn posts focus on practical HubSpot and RevOps work: building automation, making CRM data useful, and testing where AI agents help.
One implementation story shows an agent misclassifying a private equity firm’s technology roles and explains the rules added to correct it (Agent Builder example).
Another describes a lightweight Slack alert for high-intent leads that might otherwise go unnoticed (alerting example).
These examples make the content a plausible fit for campaigns about HubSpot operations, CRM automation, or applied AI.
The LinkedIn profile showed 14,760 followers on 6 October 2026.
Ask for audience data on roles, seniority, geography, and company size before assessing reach among those groups.
Pasha Irshad's posts describe client engagements, not a sponsorship program or published partnership menu.
Read more
One client, a managed IT services company, received a custom company-research agent built and debugged inside HubSpot's Agent Builder.
For another client, company-to-event associations were written back through HubSpot's API using a tested Python process.
The posts also describe one collaboration: an Unbound 2026 happy hour and ping pong event co-hosted with Kamil Rextin of 42 Agency in exchange for a large block of GTM and HubSpot engineering work, rather than a paid placement.
Brands considering a sponsorship should ask whether offers are available, what they cost and include, and how to inquire; the posts provide no rate card, partner policy, or official inquiry route.
Co-hosted event collaboration
Buyer requirement
Pasha Irshad's posts describe one Unbound 2026 happy hour with 42 Agency: the host provided the co-host slot, and they committed a substantial block of GTM and HubSpot engineering work in return. A brand proposing something similar should expect to trade event hosting or co-marketing space for engineering time and agree on the work's scope before the event.
A HubSpot implementation consultant wants to assess Pasha Irshad for a brief proposing expertise in building custom automation solutions. The client, a managed IT services provider, needs to research target companies, classify IT posture, and package the method as a repeatable offer.
Read more
Reasons
Irshad describes the job in detail: a managed IT services client, an agent built inside HubSpot's Agent Builder, the manual research it replaced, the first private equity test that misclassified technology roles as an internal IT department, the rule change that fixed it, the rerun, and the Tier 1A account list the agent now runs across.
The same posts show the scoping discipline a buyer would want before paying for agent work: define the output first, keep deterministic rules in workflows, calculated properties, custom code, or the API, and reserve the Data Agent for interpretation-heavy work such as summarizing account research, interpreting buyer intent, identifying pain points, or turning unstructured CRM activity into a usable insight.
There is working proof at the API layer too: a tested Python process creates company-to-event associations that segments and workflows would not produce, with 16 contacts resolving to 10 unique companies, eight new associations created, two existing ones preserved, and all 10 verified.
A lightweight Slack alerting build shows a second implementation pattern for the same buyer: use tools already owned, skip another dashboard, and push a daily signal about high-intent leads to where the team already works.
The background supports the pitch: ten years running demand generation as a VP of digital, followed by a move into RevOps and HubSpot because the smartest marketers were constrained by overcomplicated, under-resourced, and overfunded stacks.
Main risk
The strongest evidence comes from short-form LinkedIn posts, not transcripts, long-form articles, or code repositories. Claims about the underlying agents, scripts, and integrations rely on Irshad's descriptions, so ask for a build walkthrough or working files before treating the posts as proof of delivered code.
Verdict
Strong fit on the evidence reviewed.
When to choose a specialist
A horizontal B2B or consumer app wants a paid creator sponsorship for a broad professional audience: a fixed package, a rate card, audience demographic guarantees, and a message that AI agents can run marketing operations with little manual CRM context.
Read more
Reasons
Irshad's posts do not mention a sponsorship program, rate card, media kit, or partner policy. The one disclosed collaboration was a co-hosted Unbound happy hour and ping pong event with 42 Agency, arranged in exchange for a large block of GTM and HubSpot engineering work rather than a paid placement. Ask whether sponsorship offers exist, what they cost, and what they include.
Irshad's posts provide no audience demographics, seniority, geography, company-size, impressions, or engagement data. The LinkedIn follower count recorded on 6 October 2026 was 14,760. Ask for audience data before committing to a broad-audience or demographic guarantee.
The published material is narrow and technical: HubSpot agent builds, Data Agent credit scoping, API association work, Slack alerting, and Unbound product commentary. A brief that needs general-consumer reach or lifestyle appeal does not match the demonstrated subject matter.
Irshad's stated through-line is context-first: without context, AI will not be the answer. A campaign built on autonomous AI running marketing without maintained CRM context would conflict with that position.
Ask whether sponsorship offers exist, what they cost, and what they include; the posts provide no media kit, official partnership inquiry route, or examples of paid sponsored formats.
Verdict
Poor fit on the evidence reviewed.
A worked campaign example
Angle
Test the agent before you trust it. Use one wrong answer to sell the build.
Brief
A HubSpot implementation partner wants a specialist to propose, build, test, and hand over a custom account-research agent for a managed IT services client, then package the method as a reusable offer.
Proof points
The Agent Builder case describes a manual process that took 15 to 20 minutes per account and drew on LinkedIn, careers pages, company sites, news, and sometimes IRS Form 990 filings for nonprofits.
The first private equity test returned an internal IT team estimate of about 10 people—and it was wrong. The fix was an explicit operational-IT rule plus an entity-type check that identifies private equity and venture capital firms first.
The rerun produced the expected result, and the agent now runs across a full Tier 1A account list, sorting companies into internal team, outsourced with a confirmed vendor, outsourced with an unknown vendor, or unclear.
The API association build shows a second reusable pattern: 16 contacts resolved to 10 unique companies, creating eight new company-to-event associations, preserving two existing ones, and verifying all 10 without duplicates.
Success measure
Before committing to a full account list, a buyer can review the agent's first failure, the rule that fixed it, the rerun result, and the credit cost.
Worked campaign plan
Start with the output. Separate deterministic fields such as employee count, revenue, and account tier from probabilistic fields such as why an account may need managed IT, cybersecurity, cloud, or help desk support. Put deterministic rules in workflows, calculated properties, custom code, or the API. Reserve the paid agent for interpretation-heavy work.
Choose a test set of 10 to 20 named accounts, including at least one private equity or venture capital firm, one nonprofit, one incumbent MSP situation, and one account with unclear IT posture.
Build the agent in HubSpot's Agent Builder. For investment firms, add an entity-type check and count only operational IT roles: IT directors, systems administrators, infrastructure engineers, and help desk. Exclude data science, AI, analytics, and technology titles inside a center of excellence or an investment firm's operating bench.
Run the first pass, capture the misclassification, then rerun after the rule change. Publish a redacted before-and-after as the campaign proof point to show the build, failure mode, and repair.
Attach a credit-spend estimate before extending the agent across a Tier 1A list. For nonprofits, include the IRS Form 990 path to identify the actual IT vendor and annual spend.
Hand over the rule set, tested account list, and short runbook. If the client needs a CRM relationship HubSpot will not build natively, make the API association route a separate workstream rather than forcing it into a workflow.
If the buyer needs help deciding whether HubSpot AI credits are worth spending and how to sequence a HubSpot revenue-process rollout, Etienne Turner is the stronger alternative.
Read more
The posts price specific AI actions and prompt buyers to weigh the task's cost in human time against the AI action's cost, consider what changes at scale, and decide whether the quality is good enough to trust. The Revenue Hub post offers a rollout approach for the same decision: start with clean data, pick one use case, design around the contracts object, and land one process before expanding.
Pasha Irshad's posts focus on advice and budgets, not a completed custom agent build for a managed IT services account-research job. Ask them for the build, tested Agent Builder rules, API association workaround, and runbook for classifying IT posture.
If the buyer wants executive-level context on where HubSpot is taking Marketing Studio, self-updating CRM records, and Breeze Assistant, Yamini Rangan is the better alternative.
Read more
One post explains how those product areas are meant to reduce manual CRM work and run campaigns from a single goal.
The other sets a useful boundary for custom automation: coding has become easier, but integrations, domain depth, and trust are still hard, so a vibe-coded CRM is not the same as a governed platform.
That helps a client decide whether to build custom automation at all.
Pasha Irshad's posts focus on product strategy and platform direction, not implementation instructions for a managed IT services research workflow.
Ask them for evidence covering the first failed private equity test, credit scoping, IRS Form 990 handling, and API association steps.
Use their posts as context for the build-versus-buy decision, not as proof of delivered custom automation.
If the main risk is bad CRM data, Corrina Owens makes the stronger case for fixing the foundation before adding agents.
Read more
The posts argue that agents running on duplicate, unenriched, or inconsistently owned records will route to the wrong owner, trigger outreach from stale data, and create faster cleanup work.
The recommended order is concrete: waterfall enrichment, identity resolution and ownership, routing, scheduling, workflows, and a unified data layer so every agent sees the same customer context.
That fits a managed IT services provider whose account research depends on reliable company and contact data.
Pasha Irshad's posts focus on data infrastructure and a unified data layer, not on building a HubSpot Agent Builder classification agent for IT posture.
Ask them for the tested rule that separates operational IT roles from technology titles in a private equity firm, the nonprofit Form 990 path, and the API method for creating company-to-event associations.
Use Corrina Owens when data readiness is the blocker and Pasha when the blocker is the custom build, test, and handoff.
Pasha Irshad's eight full LinkedIn posts describe a custom HubSpot agent build for a managed IT services client, Data Agent credit and use-case guidance, a Marketing Events API workaround, a daily Slack alert script, Unbound 2026 product commentary, a Conversational Enrichment beta note, a demand-generation-to-RevOps background post, and an event co-hosting post.
The posts do not give individual post dates.
The posts here do not include transcripts, long-form articles, code repositories, sponsorship terms, rate cards, audience demographics, or a dated publication history, so claims about the agents, scripts, and integrations rest on the posts' descriptions rather than underlying files.
Missing material is noted where it affects a decision, and the posts do not describe a sponsorship offer beyond the one event collaboration.
