Shivvani Seyth

An agent has no privileged access to what happened in a business; it reads what the systems recorded.

• A CRM that stores only the current deal stage cannot show where deals actually die.
• An original-source field that anyone can edit makes every attributed answer consistent and wrong.
• In this reading, agents do not fix fragmented systems; they scale them.
• Seyth also challenges the standard enterprise answer: McKinsey's recommendation of named owners and stewards for each data domain, plus reusable data products, is sound but written for organisations with the headcount to staff it.
• The smaller alternative is four moves that need no agent at all.
• None sets one up, but all set the ceiling on what one could do later.
• The supporting numbers come from other sources, named as such: Gartner projects that agents will outnumber sellers ten to one by 2028, while fewer than 40% of sellers report productivity gains—a pattern Gartner calls agent sprawl.

One named person owns account and contact data—not a committee.

• Stage history and original source are captured automatically and locked, because nobody backfills them.
• One definition of an account holds across CRM, billing and the board pack.
• Meeting outcomes go into structured fields, not free text nobody reads twice.
• Seyth also names the layer that rarely comes with an implementation: reporting solid enough to answer a management question without an analyst building it by hand.
• Without that layer, an agent does not replace the analyst; it adds a licence fee on top of one.

The shift in treatment is visible across the corpus.

• An earlier CRM market post lists AI-driven CRM as one of four trends to watch, without conditions.
• The agentic AI post takes the opposite tack: tool choice comes down to whether the underlying data can be trusted.

Read original · Original 2 · Original 3

A generated ICP describes who you look like you should sell to.

• A company's own pipeline shows who actually bought, stayed and paid on time.
• Once two years of pipeline history exists, “no customer data required” stops being a feature.
• A website scan cannot see win, loss and churn cohorts: the segment that closes fastest is often the one that leaves in month seven, and losses carry more signal than wins but never show up on a website.

The builder reads the site, researches comparables and buying signals, weighs each segment against its case, urgency, budget and reachability, then returns a profile with citations and confidence levels.

• For a company with no pipeline history, that is a week of work in about four minutes.
• It cannot answer three questions: which segment renews, which deals take three times the effort at the same deal size, and which perfect-fit accounts were lost.
• The practical sequence is to run the builder in week one, test its output against closed-won, closed-lost and churn cohorts in week two, and trust the data when the two disagree.

Read original

Seyth argues that putting every account on the same nurture track, cadence and scoring logic—differentiated at best by industry and headcount—is infrastructure for a world where demand is evenly distributed.

• The resulting gap gets renamed a conversion problem.
• Three structural findings from a sweep of 2026 research support the argument: segmentation has overtaken measurement as teams' top database priority, contact data decays as people change roles, and buying groups are large enough that a lead-centric schema can report on them but not serve them.
• The closing point is why none of this gets done: restructuring around accounts takes a quarter and ships nothing, yet it is what makes every later campaign work.

Seyth cites specific figures from the research: 73% of teams now name segmentation their top database priority, up from measurement two years earlier, reflecting a shift from analysing segments to designing for them; B2B contacts change roles roughly every three years, leaving a database about 80% obsolete within four; and buying groups include 10 to 11 stakeholders. The figures come from 6sense, ZoomInfo, Sopro, and Demand Gen Report's benchmark survey, not Seyth's own research.

Read original

Most GTM leaders describe their teams as mostly or completely aligned, but the details that determine outcomes sit in lead handoffs, account targeting and what counts as a qualified lead.

• Much of the reported alignment turns out to exist on paper.
• Seyth's point is that alignment is not a values statement but a line in the accounts that most companies never audit.
• The supporting cost estimates are attributed to third parties: an IDC estimate of more than 10% of annual revenue leaking each year, and a HubSpot 2024 finding that sales teams in aligned organisations are 103% more likely to exceed their goals.

The Mutiny survey of 500 GTM leaders supplies the 70% alignment figure. The post closes by asking readers where their own gap lies: definitions, handoffs, or nobody owning the seam between the two teams.

Read original

They argue the omission was a misread, not a footnote: an entire cluster of the 37 sales skills in the launch concerns the installed base—account plans, customer health, renewals and expansion already sitting in accounts won two years ago.

• The focus is managing, not prospecting.
• The underlying reason that work is hard is the same one Seyth applies elsewhere: the customer data was never missing; it was unassembled.
• Assembling it cost about half a day per account, and an account manager has sixty.

The post covers four account-management problems the launch addresses: the risk spotted nine months too late, the expansion never qualified, the stakeholder map that went stale in 2024, and the tax paid around every customer conversation. Seyth also writes about what gets harder rather than easier, and says that section matters more than the four.

Read original

With the Irish minimum wage rising to 14.15 euros an hour, about 48% in eight years, agencies running field acquisition teams at or near that floor face a near-doubling of direct labour costs.

• The standard rate card no longer fits the brand procurement benchmark, and Seyth argues that accepting a labour-hour bid at thin margin is a faster route to a failed engagement than declining the work.
• The proposed fix is a different commercial structure, not a discount: retention-vested commission, with a small fixed fee for overhead plus a variable payout against milestones the agency cannot fake, as the default for door-to-door bids in subscription businesses.
• Seyth argues that the incentive alignment makes the contract worth signing.

The published argument sits behind a linked long piece that is not described here, so the summary Seyth gives in the post itself includes the wage figures and the rate-card and procurement points.

It is the only post in the corpus with a geographic anchor, on agency labour economics, and arguing about contract structure rather than data or team design.

Read original

A strong starting point for teams assessing AI sales tools or planning a data-readiness discussion.

Seyth argues that AI agents cannot fix gaps in the systems they rely on; they can only work with what those systems record.

• A CRM that keeps only the current deal stage cannot show where deals actually die, while an editable original-source field makes attribution unreliable.
• The post lays out four practical foundations: assign an owner to account and contact data; capture and lock stage history and original source; agree on one account definition across systems; and record meeting outcomes in structured fields.
• Its key distinction: buying AI tools is different from preparing the data model that determines what those tools can do.

For brands planning a campaign

Could anchor a data-readiness webinar or diagnostic for revenue teams considering AI tools, using the four foundations as a practical checklist.

Seyth tests HubSpot for Startups’ free ICP Builder on their own site.

• The tool uses website information and market signals to propose a customer profile, which they see as useful for a pre-revenue company.
• But once a business has pipeline history, a website-based profile cannot show which segments renew, which deals take far more effort than their value suggests, or what lost and churned customers have in common.
• Their practical sequence: use the tool for an initial hypothesis, then compare it with closed-won, closed-lost, and churn data.
• If the generated profile and customer outcomes disagree, trust the outcomes.

For brands planning a campaign

The week-one hypothesis and week-two validation sequence could shape an ICP workshop or audit, with win, loss, and churn cohorts as core deliverables.

Helpful for deciding whether the next investment should be in campaigns or the account data behind them.

Starting with the claim that only 5–7% of target accounts are in-market in a given quarter, Seyth asks why so much demand-generation infrastructure treats accounts as interchangeable.

• They connect segmentation to three structural issues: contact data decays as people change roles, buying groups include many stakeholders, and lead-centric databases make account-level targeting difficult.
• Rebuilding data around accounts takes time and may not look like a campaign win, but it can make later campaigns more effective.

For brands planning a campaign

Could frame a segmentation or database review around the practical gap between lead-based reporting and account-based targeting.

Seyth reads a launch of 37 sales skills through the work of account managers, arguing that the installed base deserves more attention than prospecting.

• They point to account plans, customer health, renewals, and expansion as areas the launch could support, then identify familiar operational problems: late warning of risk, missed expansion opportunities, and stakeholder maps that have gone stale.
• The common thread is not a lack of customer data, but the time it takes to assemble scattered information across a large portfolio of accounts.

For brands planning a campaign

A useful model for product content aimed at customer success or account-management teams: connect a feature set to concrete workflows and clarify what work remains difficult.

A useful starting point for an internal review of shared definitions and handoffs.

Seyth contrasts leaders’ reported confidence in sales and marketing alignment with the day-to-day mechanics behind it: lead handoffs, account targeting, and the definition of a qualified lead.

• The post argues that alignment is more than a shared statement of intent; it has operational consequences that can be examined in revenue terms.
• It asks where the gap sits—in definitions, handoffs, or team responsibilities—rather than treating alignment as an abstract culture issue.

For brands planning a campaign

Could introduce an alignment diagnostic focused on qualification rules, handoffs, and account targeting rather than a broad survey of team sentiment.

In the agentic AI post, Shivvani Seyth points to customer, pricing, transaction, and interaction data fragmented across systems that disagree.

• They propose steps that work without an agent: one named person owning account and contact data; automatic, locked capture of stage history and original source; one definition of an account across CRM, billing, and board reporting; and meeting outcomes captured as structured fields.
• The Claudeforce post makes the same point about account management: renewal, health, expansion, and stakeholder data existed but remained unassembled because assembling it cost about half a day per account.

Starts with a named third-party report or product launch, then translates it into steps a small revenue team can take without adding headcount.

For brands planning a campaign

For campaigns on AI in sales, CRM migration, or revenue data, lead with data readiness, not the tool; a concrete data-model fix can provide the content hook.

Examples: Source 1 · Source 2 · Source 3

The misaligned slice is estimated to leak 10%+ of annual revenue (an IDC estimate).

• Sales teams in aligned organisations are 103% more likely to exceed goals (HubSpot 2024 Sales Trends Report).
• The post asks readers where the real gap lies: definitions, handoffs, or nobody owning the seam.

For brands planning a campaign

Suits campaigns on alignment, handoff design, or lead definitions, with a research-backed frame that treats alignment as an audited number, not a values statement.

Examples: Source 1

Seyth describes what it does — reads the website, researches comparables and buying signals, weighs segments on evidence, urgency, budget, and reachability, and returns a profile with citations and confidence levels — and where it falls short once pipeline history exists: which segment renews after first purchase, which deals cost three times the effort at the same deal size, and which "perfect fit" accounts were lost. The recommendation: run the builder in week one, then test its output against closed-won, closed-lost, and churn cohorts, and let the data win when the two disagree.

A hands-on tool test, followed by a direct comparison with the author's own pipeline outcomes.

For brands planning a campaign

A strong fit for ICP, segmentation, or sales-data content: the tested-tool framing gives a brand a specific example to build around.

Examples: Source 1

It highlights three findings from 2026 research: 73% of teams now name segmentation their top database priority, up from measurement two years earlier; B2B contacts change roles roughly every three years, leaving a database about 80% obsolete inside four; and buying groups run 10–11 stakeholders, which a lead-centric schema can report on but cannot serve. The post argues that restructuring around accounts takes a quarter and ships nothing, so it never gets prioritised.

For brands planning a campaign

Well suited to demand-generation, database-hygiene, and buying-group campaigns that propose structural change rather than another campaign.

Examples: Source 1

The core claim: an agent reads only what systems record, so an editable original-source field makes every attribution answer "fluent, consistent, wrong." The post also covers the post-go-live reporting gap: implementations are configured competently but rarely arrive with reporting solid enough to answer a management question without an analyst building it by hand.

The claim that an agent without a reporting layer "adds a licence fee on top of" an analyst is stated as an opinion; campaigns reusing it should attribute it.

For brands planning a campaign

For AI-in-sales and agentic-tooling campaigns, the differentiator is the data and reporting prerequisite, not the agent itself — a fit for this audience's skepticism about launch hype.

Examples: Source 1

It argues for restructuring the commercial model rather than discounting, proposing retention-vested commission — a small fixed fee for overhead plus variable payout against milestones the agency cannot fake — as a default for door-to-door bids in subscription businesses. Incentive alignment, it argues, makes the contract worth signing.

This post takes up agency labour economics for field acquisition teams, a distinct strand from the CRM and RevOps topics.

For brands planning a campaign

Applies to field marketing, agency procurement, or subscription customer-acquisition campaigns, particularly in Irish and European markets; a narrower fit than the RevOps topics.

Examples: Source 1

Third-party sources are named where relevant — McKinsey, Gartner, 6sense, HubSpot, IDC, Mutiny — and several posts point to a longer written piece in the comments. One pairs the analysis with a first-person tool test; the CRM market overview lists four trends and invites readers to share how their own strategy is changing.

For brands planning a campaign

Expect short-form posts built on cited third-party research that point to a longer article; budget for both the post and the linked piece. The closing question offers a natural comment prompt.

Examples: Source 1 · Source 2 · Source 3 · Source 4 · Source 5 · Source 6 · Source 7

The post does not state their relationship to HubSpot; ask for clarification. HubSpot's 2024 Sales Trends Report is also cited in a separate post on sales and marketing alignment.

View example

Salesforce is named through the product and a hashtag; ask whether there is a commercial tie.

View example

View example

View example

View example

Newman's Slack post and Seyth's agentic AI post also cite the same Gartner projection: agents will outnumber sellers ten to one by 2028, while fewer than 40% of sellers report productivity gains.

Newman works from inside a working stack, showing a repeatable Slack-and-CRM briefing built prompt by prompt and listing criteria for scoring AI vendors before a demo. Seyth takes a step back, arguing that the underlying data model and account definitions must be in place before teams can trust the tooling.

Read original · Original 2

Tatulea's account hierarchy examples and Seyth's point about editable original-source fields make the same case from different angles: fluent answers built on wrong records.

Tatulea writes from an ABM execution and vendor-product perspective, including the launch of a context graph for go-to-market teams, and organises the argument around the four layers of an ABM play. Seyth reads third-party research from the outside, testing its claims against what a small revenue team can actually staff.

Read original · Original 2

Spira's case that the first revenue operations hire should own definitions, architecture and process before an analyst builds any report echoes Seyth's argument that data models must precede AI.

Spira's material focuses on deal governance and hiring sequence, including a documented collaboration with a deal-management vendor on audit trails and approval history. Seyth writes about data readiness for AI and demand infrastructure, keeping the commercial work in the background rather than inside the post.

Read original · Original 2

Newman's posts document a working pattern for exactly that.

• One prompt summarizes the past seven days of sales activity across channels into key decisions, deals mentioned, and a prioritized follow-up list with the people to contact.
• A second turns last quarter's conversations into a growth briefing canvas covering overview, channel summaries, enablers, blockers, and follow-ups.
• The prompt is then saved as a reusable Slackbot skill, scheduled for the last Friday of each quarter, and delivered with a reminder to prep for the quarterly review.
• Alongside that, Newman publishes a five-point filter for buying AI at all — does it fit the team's existing workflow, does it plug into the CRM, are the data boundaries clear, does it improve over time, and what is the measurable return — backed by Gartner's figure of agents outnumbering sellers ten to one by 2028 with fewer than 40% of sellers reporting productivity gains, and by MIT's finding that 95% of organizations are getting no return from generative AI.
• If a revenue team needs to use AI it has bought inside Slack and the CRM, this material addresses the task more directly than a data-foundation diagnostic.

Choosing Newman over Seyth means giving up the layer beneath the tools.

• Neither Newman post addresses account identity and entity resolution, ICP testing against closed-won, closed-lost, and churn cohorts, or how a sales and marketing team defines a qualified lead.
• The engagement moves quickly through workflows — configure the prompt, save the skill, schedule it, score the vendors — but leaves unanswered the questions that determine whether those workflows produce reliable results: who owns account and contact data, whether stage history is locked, and whether systems share one definition of an account.
• Newman's two posts offer single-recipe pieces rather than a body of diagnostic work, so the buyer gets a practitioner who can build a working automation, not someone with a published method for finding errors in revenue data.

Examples: Source 1 · Source 2

The posts name that failure with concrete examples: Google and Waymo sitting as two separate accounts with no link to Alphabet, a hierarchy that never updates when one company acquires another, and a marketing system that still places an executive at a company they left.

• Tatulea states the consequence in buying terms — right person wrong account, right account wrong person, right person and account but wrong message — and argues the next era of go-to-market goes to whoever gives the agent the best understanding of the world it operates in, not the most tools.
• The companion post lays out the four layers any account-based motion needs (a named list of target accounts, the named buyers at each, signals proven to predict conversion, and accumulated context on each account) and why they fail: each layer lives in a different system, revenue operations stitches them by hand, the stitching goes stale, and an agent chains the steps so one wrong field becomes the assumption everything later inherits.
• If the need is to fix a CRM that cannot tell which company is being sold to before pointing an agent at it, Tatulea is closer to the work than a diagnostic that starts at definitions.

Part of the account-view post is built around a specific context-graph product launch, so some of the argument reads as platform positioning rather than independent diagnosis.

• Buyers should separate the problem description from the product claim and score the vendor like any other.
• The coverage spans two posts on data-platform maturity; choosing Newman means giving up Seyth's ICP work (generating a profile, then testing it against renewal, win rate, and churn cohorts), the sales-and-marketing alignment audit, and the content-engagement shape of a research-led post driving to a longer article.
• The engagement also leans toward sourcing and standing up a data layer, a different purchase from auditing what the client already owns.

Examples: Source 1 · Source 2

One post makes the case that a signed deal should be reconstructable — which quote was final, who approved the discount, when the terms changed, whether the CRM matches what the customer signed — and reframes the audit scramble as a revenue governance problem rather than paperwork.

• It argues that version control and approval history belong to the deal itself so no one has to remember where anything lives.
• The other post argues the first revenue operations hire should not be an analyst, and gives the order of operations that hire has to own: definitions first (what counts as a lead, an opportunity, a stage, a qualified deal, a commit), then architecture that enforces them, then process, then measurement, then automation and AI on a system that is already true — with a concrete interview test of handing a candidate your stage definitions and asking which they would delete.
• If the need is quote-to-contract traceability, audit readiness, or choosing and structuring the first revenue operations owner, these posts address it directly.
• Seyth's posts do not cover deal governance or hiring sequence.

One post is built around a named product and reads as vendor-adjacent, so the governance argument comes with a tool attached; buyers should judge the framing separately from the platform.

• The two posts offer no demand-generation or ICP material — nothing on buying groups of ten or eleven stakeholders, database decay, or segmentation as a database problem.
• A client whose weakness is on the demand side rather than the deal side should ask for a second engagement covering those topics.
• Spira's posts are short, punchy, confessional pieces written to prompt comments rather than the research-led post-to-long-article format a software company could commission as thought leadership.

Examples: Source 1 · Source 2