GTM & RevOps
01SaaStr
Databricks held SDR headcount flat as AE count grew
Databricks automated outbound research and personalization, kept SDR headcount flat while adding AEs, and moved reply rates from 1-2% to 6-10%.
What happened
- CRO Ron Gabrisko kept SDR headcount roughly flat while continuing to add AEs and solutions architects.
- Outbound personalized down to the individual lifted reply rates from 1-2% to 6-10%, replacing 5,000-contact blasts.
- At Monaco, one SDR books 400 outbound meetings a month with research, signals and messaging automated.
- Databricks keeps the SDR team as a training pipeline, since most of its AEs are promoted from it.
Why it matters
The growth lever is no longer SDR headcount but the signal and personalization layer sitting behind each rep. Pipeline coverage can scale without the comp line scaling with it.
The JayOh take →
Before approving another SDR req, we would instrument the signal layer first - account research, trigger detection, per-contact message assembly - and hold headcount for a quarter to see what per-rep output actually does.
02SaaStr
AE candidates build a live demo in the interview
Every Databricks AE candidate builds a product demo on the spot and pitches a named account at CEO level.
What happened
- Candidates pick an account, give an industry-specific CEO-level pitch, then build a demo live for the panel.
- After hiring, reps are certified on the product and the pitch, with recertification running through the year.
- Gabrisko targets 60% of rep time in front of customers, confining prep and admin to Monday and Friday.
- Demos run on data resembling the buyer's industry; Monaco replaced the CSM role with forward deployed AEs.
Why it matters
If sellers are expected to build, the hiring loop and enablement stack have to test building rather than storytelling. Most loops still screen for deal stories and role plays.
The JayOh take →
We would add a build task to the AE interview loop and stand up a demo environment seeded with industry-shaped sample data, so a rep can show a working version before the buyer commits.
03PPC Land
Google ad growth slows as AI Mode takes the space
Alphabet's ad line grew just 10.5% to about $82B as AI Mode compressed search ad slots, and shares fell 11%.
What happened
- Alphabet posted $127.6B in quarterly revenue, up 25% year over year, with Cloud up 103%.
- Advertising rose 10.5% to roughly $82B, the line investors reacted to.
- Shares fell about 7% after hours and later by more than 11%.
- Google said ads in AI Mode would now reach 10 more markets.
Why it matters
Paid search inventory is being rewritten by AI answers. Demand gen plans built on stable search volume and CPC behavior are planning against a surface that is still moving.
The JayOh take →
We would stop treating AEO as a content side project. Model-answer visibility now competes directly with the paid slot, and attribution needs to capture AI-answer assisted sessions before budgets shift mid-quarter.
04The AI Show
Attackers ran agents against 400+ financial institutions
A criminal group used AI agents to plan attacks and harvest credentials across more than 400 Latin American financial institutions.
What happened
- Researchers attributed the campaign, dubbed Cyber Cartel, to coordinated agent use for planning and credential collection.
- Anthropic revoked the attacker's access, and separate campaigns targeted South Korean financial firms.
- Two PHP vulnerabilities, present in the code for 14 years, were patched October 21 after exploit attempts on October 17.
- Cloudflare logged thousands of exploit attempts before public disclosure and millions within twelve hours.
Why it matters
Every agent wired into a CRM holds credentials and write access. The attack surface of a revenue stack now includes the automation layer, not only the people using it.
The JayOh take →
We would inventory every agent and integration holding write access to the CRM, scope their tokens to the minimum object set, and put a revocation path in the runbook before adding another one.
05Latent Space
Frontier models are picking the default B2B stack
One backend swept first-choice recommendations across every frontier model tested, making model defaults a distribution channel.
What happened
- Latent Space's AEO tracker found an almost complete sweep of first-choice app backend recommendations across frontier models.
- The same company reached a $10B valuation as an open-source, fully remote business.
- Familiarity to the models explains part of it; the rest is the work of making generated apps actually run.
Why it matters
When a buyer asks an assistant what to use, the model's default answer is the shortlist. That default is earned through documentation and machine-readable surface area, not ad spend.
The JayOh take →
We would run AEO as a product-marketing workstream: publish docs and integration examples a model can quote, then track first-choice placement by category the same way we track rankings.
06Kyle Poyar
B2B newsletter ad inventory is sold out for months
Niche B2B creators are turning away advertisers, and one newsletter ad network has paid out more than $37M to writers.
What happened
- Poyar describes creators with inventory sold out months ahead and multiple partnership requests a day.
- beehiiv's newsletter ad network has paid more than $37M to writers.
- B2B creators here are practitioners with audiences in the thousands, not mass-reach social influencers.
- Poyar sells one partner per theme per quarter to keep ads adjacent to the topics buyers care about.
Why it matters
Attention in B2B is consolidating around individual practitioners. Buying it now takes lead time and a placement strategy rather than a quarterly media test.
The JayOh take →
For a narrow ICP, we would book two or three niche newsletters a quarter ahead and track them as pipeline sources in the CRM rather than filing them under brand spend.
Platform & Tool Releases
07HubSpot
HubSpot enforces CRM validation rules on API writes
From API version 2026-09 admin validation rules apply to API writes, and an October 27 note narrows what is actually enforced.
What happened
- Conditional required properties now return a 400 with MISSING_CONDITIONAL_REQUIRED_PROPERTY on API writes.
- Properties marked required in Create Record settings are enforced on POST calls.
- An October 27 correction states required associations are enforced in the app UI only, not on public API write paths.
- Success responses now include a warnings array and error message shapes have changed.
Why it matters
Integrations that wrote records around UI-only requirements will start failing, and any team that assumed required associations were enforced everywhere has a silent data-quality gap.
The JayOh take →
We would pull property and Create Record settings per portal, diff them against what each integration writes, and handle the new 400s explicitly instead of letting a sync retry quietly in the background.
08HubSpot
Deal progression ranks next actions from call history
A public-beta card on the deal record proposes property updates and next best actions drawn from logged calls and meetings.
What happened
- Deal progression sits on the deal record's Catch-up tab for Sales or Service Hub Professional and Enterprise.
- It ranks next best actions across every call and meeting on the deal, not only the most recent one.
- Reps accept or dismiss suggested property updates and can draft emails, tasks or meetings from recommendations.
- A configured methodology such as MEDDIC or BANT is used as context, and a Super Admin must enroll in the beta.
Why it matters
This shifts deal hygiene from rep discipline to a suggestion queue, but only works where the methodology fields and stage definitions have actually been configured.
The JayOh take →
We would configure the methodology fields and tighten stage exit criteria before enrolling a portal, otherwise the agent starts recommending actions against definitions nobody on the team agreed to.
09Salesforce
Salesforce ships an agent kill switch and spend caps
Agent Kill Switch and FinOps controls went GA on October 27, with cost policies cutting token usage up to 75% in tests.
What happened
- Agent Kill Switch, which pulls an agent offline, became generally available October 27, 2026.
- The automatically triggered version is slated for general availability in January 2027, alongside Agent Evals.
- FinOps Dashboard, Model Wallet hard limits and Cost Optimization Policies are generally available now.
- Salesforce says cost policies cut token usage by up to 75% in test scenarios via caching and payload compression.
Why it matters
Agent platforms are converging on the same two controls: stop it, and cap it. Both belong in a rollout plan from day one rather than after the first surprise invoice.
The JayOh take →
We would require a documented kill path and a per-agent spend ceiling before an agent touches production data, with a named owner for the threshold - the control is worthless if nobody sets it.
10The Rundown AI
Gemini 4 Carbon tops benchmarks at half the price
Google's Carbon leads 15 of 17 benchmarks at $2/$10 per million tokens, but access is limited to select cyber defenders.
What happened
- Carbon scores 59 on Artificial Analysis' Intelligence Index, nine points above Google's Argon from four weeks earlier.
- Its widest margins came on long-input reasoning, security work, workflow automation and math.
- Intro API pricing is $2 in and $10 out per million tokens, doubling once the discount period ends.
- No date was given for wider access; Argon reached broader availability roughly two weeks after its debut.
Why it matters
Frontier pricing keeps falling while capability rises, which changes what is affordable to run across a full CRM dataset rather than a sample of it.
The JayOh take →
We would re-cost any agent workload shelved on price grounds this quarter. The per-task economics of enrichment, summarization and research now move every few weeks.
Catch up on past briefings in JayOh University.
Turn these signals into your GTM roadmap.
We engineer the CRM, RevOps and AI systems behind B2B revenue teams.
Talk to JayOh