Simply Broken
Exploring / Avero: The Future of CRM and Its Agencies / v1
Version 1 · 1 September 2026 · the wide map

The future of CRM, and of the agencies that build it

Ten parallel research passes across the CRM vendors, the AI-native challengers, and the implementation agencies that serve small and mid-size companies. This page is a learning and steering document for Avero: what is actually shipping, what is collapsing, who is winning with evidence, and eighteen published frameworks laid side by side.

How to read this page. Every claim that matters carries a confidence tag: High means multiple dated primary sources agree, Medium means a single strong source or corroborated direction, Low means one thin source, a vendor talking its own book, or content farms citing each other. Where the research found nothing, the page says so instead of smoothing it over. The eleven questions this page answers are Amit's own, kept verbatim at the foot of the page.
The classic CRM agency 1 2 3 4
The squeeze, and the ladder out. The numbers are explained below; the whole page is the evidence for this one drawing.
  1. Vendors absorb the work from below. Enrichment became a native feature when HubSpot bought Clearbit. Migration has a one-hour AI agent. HubSpot's own partner playbook tells agencies to stop selling basic implementation.
  2. AI compresses the hours from the side. The billable-hour assumption, that time and value move together, is breaking. Fixed fee per milestone already dominates among named HubSpot partners.
  3. Buyers move upstream from above. 51 percent of B2B software buyers now start research in an AI chatbot, and ranking on Google no longer predicts being cited by the AI.
  4. The ladder out runs through judgment. Architecture, data foundations, agent governance and origination of demand. The firms climbing it are named on this page; the ones standing still are also named.

Question 1What is happening in CRM right now

Eight findings survived cross-checking across the ten research passes. Each one opens to its evidence.

01

The agency is not dying. The generalist is. High

Demand for Salesforce consultants grew 10 percent globally and 72 percent in the US in 2025, while every credible source describes the same bifurcation: a small premium tier selling AI-ready architecture, and a large tier selling implementation with shrinking margins.

The evidence, and the three-phase history it fits.

Salesforce Ben (May 2026) carries the named accounts: Joey Monroe, a Salesforce consultant, felt his job was at risk a year earlier and repositioned into AI strategy advising. The demand figures in the same piece are IDC-sourced. The conclusion is not that consultants disappear; it is that config-only and generalist roles do.

Logarithmic Perspectives frames the partner ecosystem in three phases: 2010 to 2016, implementation contractors (thin margins, low differentiation); 2016 to 2022, operations managers (managed services, embedded support); 2026 onward, strategic architects (AI-ready data foundations and integration topologies). Its line worth keeping: "Implementation and managed services are operational skills. Designing AI-ready integration architectures is a systems engineering discipline."

Adjacent but real: Salesforce's own support org shrank from roughly 9,000 to 5,000 people as Agentforce absorbed conversation volume (Benioff, September 2025). That is the vendor's staff, not agency staff, but it shows the absorption pattern working at scale.

02

Per-seat pricing is cracking, and no vendor has found the replacement. High

Salesforce shipped three different Agentforce pricing models in 18 months. HubSpot moved to outcome pricing: 50 cents per resolved conversation. Freshworks raised per-session pricing five-fold in one quarter. Microsoft went the other way and bundled Copilot down to SMB price points.

The model-by-model ledger.

  • Salesforce: $2 per conversation (Oct 2024), then Flex Credits at $500 per 100k (June 2025), then per-user at $125 to $550 a month (Q1 2026). Three models, none settled. Entry cost stays enterprise: the Service Cloud prerequisite alone is $165 per user per month.
  • HubSpot: Breeze Customer Agent moved from a flat $1 per conversation to $0.50 per resolved conversation on April 14, 2026. The only vendor pricing on success rather than activity. Prospecting Agent is $1 per recommended lead.
  • Microsoft: cut the SMB Copilot tier from $30 to $21 (December 2025, $18 with annual commitment), bundled basic Copilot into Dynamics licenses, and made Business SKUs with Copilot permanent in July 2026. The most SMB-favorable incumbent by far.
  • Freshworks: raised Freddy AI sessions from $0.10 to $0.50 in Q4 2025, a five-fold increase that suggests usage ran far above what they modeled.
  • Zoho: announced Zia Agents in July 2025 as included in Enterprise and Ultimate plans; 14 months later there is still no confirmed hard general availability. Medium
03

Both big vendors rebuilt their partner programs into origination filters. High

HubSpot now pays 20 percent for three years only on deals the partner sources; a referred-in deal earns one year. A mandatory $400 a month membership fee arrived July 2026, with a termination wave for partners who did not convert. Salesforce collapsed four tiers to two and gated the top one behind proven Agentforce delivery.

What changed, dated, and what it does to an agency's pipeline.

HubSpot: lifetime commissions became 3-year commissions in April 2023. The sold-versus-sourced split (July 2025) pays sourced deals 20 percent for 3 years and assisted deals 20 percent for 1 year. From November 2025, no shared deal record means no credit at all. From January 2027, untiered partners must show sourced deals or leave. The $400 a month membership fee became mandatory July 15, 2026; unconverted partners were terminated August 15, 2026, commissions ending immediately.

Salesforce: the FY27 program (announced March 2026) went from four tiers to two (Summit, Select), replaced 170 badges with 28 competencies weighted toward Agentforce and Data 360, and rebranded AppExchange toward AgentExchange with a $50M builders fund. At the same time it is raising partner investment: a $1B incentive target and a doubled partner fund.

The read: vendors are not killing the channel; they are consolidating it toward fewer, AI-fluent partners who bring their own demand. The directory-fed lead is now worth a third of what it was relative to a self-originated one. Waiting for vendor referrals is the strategy that just got repriced.

04

The buyer moved upstream, into the chatbot. Medium

G2's Answer Economy survey (1,076 B2B software buyers, March 2026): 51 percent start research in an AI chatbot, and chatbots are the number one shortlist influence at 54 percent, ahead of review sites and vendor sites. Ranking in Google's top ten no longer predicts AI citation: the overlap collapsed from about 75 percent to 17 to 38 percent within a year.

Why the tag says Medium, and what replaces classic SEO.

G2 sells visibility into exactly this market, so its framing serves its book; the sample is large and the direction is corroborated by independent AI-Overview mechanics (organic results under an AI Overview lose about 61 percent of click-through; zero-click searches are near 60 percent of all searches). Direction: high confidence. Magnitudes: estimates.

What the sources say replaces it: being cited by the models, which follows topical authority and third-party mentions in sources the models trust, not keyword optimization. The one documented mechanism case: Apollo, mis-described by LLMs, reportedly reached 6x AI-referred trials by treating the problem as narrative control across third-party sources. Low on that figure; it traces to a vendor roundup. And llms.txt, commonly pitched as the fix, is explicitly unsupported by data as a ranking factor.

05

The realest architectural shift is MCP: the CRM as a tool surface for agents. High

HubSpot's remote MCP server went GA on April 13, 2026. Salesforce unveiled Headless 360 at TrailblazerDX 2026 with 60 plus MCP tools. Both of the biggest vendors committed to the same protocol direction in the same year, which is the strongest durability signal in this entire research pass.

Why this matters more than the agent demos.

APIs are old; what is new is a standardized, typed, agent-discoverable tool protocol with OAuth, so any MCP-compatible agent can call CRM functions without a bespoke integration. The unit of implementation work then shifts from "build screens and automations in the vendor UI" to "define and govern the tool surface: what an agent may read, what it may write, and when it must escalate to a human." That deliverable did not exist as a line item two years ago.

The challenger cohort is building for the same thesis from the other end: Day.ai (founded by HubSpot's former CPO, $24M from Sequoia) makes passive capture the CRM's primary write path; Clay is a data orchestration layer that never claimed to be a CRM and reached $100M ARR in two years; Monaco ($35M from Founders Fund) auto-updates pipeline with no manual entry. The highest-conviction new money treats the CRM as infrastructure agents read, not a UI humans type into.

06

The hype has a body count, and the gap is governance, not capability. High

MIT NANDA: 95 percent of enterprise GenAI pilots show no measurable P&L impact. Gartner projects over 40 percent of agentic AI projects canceled by end of 2027 and names "agent washing" as widespread. Salesforce reports $1B Agentforce ARR while its own partners tell The Register they see no meaningful revenue from it.

The named failures worth remembering.

11x.ai (autonomous AI SDRs, $74M from a16z and Benchmark): TechCrunch found it listing ZoomInfo and Airtable as customers when neither was, ARR inflated by annualizing short trials, churn reported at 70 to 80 percent. The founder stepped down as CEO in May 2025.

Artisan ran the "Stop Hiring Humans" billboards in 2024; in August 2026 it retired the slogan and hired its first human BDR. A company selling "AI replaces SDRs" concluding it needs a human SDR is the cleanest tell in the whole dataset.

The adoption reality: only about 12 percent of Salesforce's customer base had any Agentforce deal, about 6 percent paid, and over 60 percent of bookings were expansions into existing customers. SaaStr's headline for practitioners: "83 percent of you haven't gotten AI SDRs to work yet." Medium on those exact figures, high on the direction.

The consequence for an agency: if 40 percent of agent projects die from cost, unclear value, or missing risk controls, the highest-margin service is not building the agent (the vendor sells that self-serve). It is the governance layer: what the agent may touch, when it escalates, and how anyone would know it is working.

07

GTM engineering is the fastest-forming new category on the agency side. High

The term barely existed before 2024. Clay went from $1M to $100M ARR in two years and was valued at $5B in January 2026. A whole "Claygency" sub-sector formed, and Clay fluency now sits next to vendor certification as a co-equal credential: RevPartners holds Elite status from HubSpot and from Clay simultaneously.

Who and what, with prices.

84 percent of surveyed GTM engineers use Clay; 96 percent among agencies (Clay's own survey of 228, so read with that in mind). Named shops: Understory, ColdIQ, SalesCaptain, Growth Engine X, OneAway (190 plus Clay tables across 50 plus engagements), demandDrive (Clay Studio Partner, May 2026). Going rates: entry Clay builds $2K to $5K a month, full-service $8K to $15K a month per client. The service bundle is outbound automation plus enrichment plus CRM hygiene sold as one thing, which is outbound tooling and CRM implementation merging into a single category.

08

Consolidation started, and recurring books price higher, with one twist. Medium

SmartBug bought Globalia (2023). Aptitude 8 bought WORQFLOW (2024). SBI Growth bought Carabiner (2024). Walker Sands bought RevPartners (June 2026). Recurring revenue commands a premium of one to two multiple turns in agency M&A. The twist: the fastest-growing agencies in one 300-firm benchmark ran more projects and fewer retainers than their peers.

Question 2The software: shipped versus announced

39 features tracked across seven incumbents, verified against dated primary sources: 25 shipped GA, the rest beta or announced only. The pattern that matters for an SMB-focused agency: Microsoft is buying the SMB market with bundling, HubSpot is the only one pricing on outcomes, and Salesforce remains an enterprise-cost proposition whatever the demo says.

SalesforceAgentforce 3, GA June 2025
Multi-agent orchestration, observability, commerce agents all GA. But SMB relevance is low: first-year cost for 30 seats runs $200K to $450K once the Service Cloud prerequisite is counted. Partners touch 70 percent of implementations.
HubSpotBreeze, GA Sept 2024
Customer Agent, Prospecting Agent, Data Agent shipped; Agent Hub in public beta July 2026. Outcome pricing since April 2026. High SMB relevance: no seat minimums, pay per result.
MicrosoftCopilot wave, GA Oct 2025 onward
Role-based Copilots bundled into Dynamics licenses at no extra cost; SMB tier cut to $21 then $18 promo; permanent Business SKUs with Copilot July 2026. The most aggressive downmarket move of any incumbent.
ZohoZia Agents, announced July 2025
25 plus prebuilt agents and a no-code Agent Studio, bundled into Enterprise ($40 per user) and Ultimate ($52). Still reads as early access 14 months later; no confirmed hard GA. Agent Studio is Ultimate-only, which creates SKU friction.
Pipedriveincremental AI, MCP June 2026
AI reports, email assistant, ChatGPT integration shipped; the full agentic experience is still pilot-only. Notable: a native MCP server in all plans since June 2026. AI is bundled, not an add-on line.
monday CRMSidekick GA Jan 2026
Credit-based AI pricing ($0.01 a credit) applies to the work platform only; monday CRM itself stays seat-based. Confusing in market, and worth knowing before recommending.
FreshworksFreddy agents, GA Q4 2025
Support-centered agents at $0.50 a session after the first free 500; Copilot for human agents at $29 a month. Positioned as uncomplicated AI for CX, not a horizontal agent framework.

The challengers, in one paragraph each worth knowing

AttioSeries B $52M, GV, Aug 2025
Flexible schema the AI reads and writes directly; 5,000 paying customers. "Self-implementing" is marketing gloss over fast manual setup with AI assist, and that gap is agency opportunity.
Clay$100M ARR, $5B valuation, Jan 2026
Not a CRM: a GTM data orchestration layer with an agent (Claygent, 1B plus executions). Kept its reputation by never claiming full autonomy. The center of the new agency category.
Day.aiSeries A $20M, Sequoia, Feb 2026
Founded by HubSpot's former CPO. Conversation-native: the CRM writes itself from email, calendar and calls. The genuine architectural inversion in the cohort. 120 customers, very young.
TwentySeries B $100M, $1B valuation, June 2026
Open-source CRM, 44K GitHub stars, 300 plus contributors. Proof that self-hosted, no-lock-in CRM can reach scale.
Monaco$35M launch, Founders Fund, Feb 2026
End-to-end autonomous revenue orchestration for early-stage startups; pipeline updates itself. Early and unproven, but the thesis is the data layer one.
Folk, Breakcold, Gorgiasvertical plays
Folk owns founder fundraising, Gorgias owns ecommerce support-plus-sales (15,000 brands), Breakcold ships MCP-native CRM for agencies at $59 a month. Verticals are outpacing new horizontals.

Questions 3 and 7The work itself: what changed, who does it

Three years ago the shape was a delivery pyramid: seniors scope, juniors configure, offshore fills the bottom. What the evidence supports now is narrower and more interesting than the headlines.

What is well evidenced

Role redefinition, not title deletion. No named agency was found deleting "solution architect" or "admin" from its org chart. What is evidenced is task content moving: AI credited with routine config, migration mechanics, documentation and first-pass QA, while humans keep architecture judgment, data governance, and the client relationship. New titles are appearing in CRM job posts: an "AI Implementation Specialist" posting built around Salesforce and HubSpot with prompt-engineering experience preferred, at $50K to $90K. High

Delivery pods, senior-led. RevPartners runs "RevOps Pods" on a fixed monthly subscription. A 2026 pattern worth watching: pods led by very senior engineers working alongside a set of AI agents running discovery, metadata review, build and testing. Some agencies now market "seniors execute, no delegation to juniors" as the premium pitch, which produces the same visible effect as AI substitution and is easy to misattribute. Medium

Timeline compression is claimed everywhere and measured nowhere. Every 2026 source asserts 6 to 8 week Professional onboardings, one markets a 42-day sprint; not one pairs that with a matched 2023 baseline. Treat compression as a consistent industry claim, not a measured fact. Low

Live contradictionOffshore versus onshore has not converged. Offshore-industry sources argue AI compounds the arbitrage: a cheaper rate times AI throughput is a lower cost per delivered unit. Aptitude 8, an Elite HubSpot partner rated 5.0 across 250 reviews, markets the exact opposite: in-house, US-based, no outsourcing, as a trust differentiator. Both positions are winning in the same market in 2026. Nearshoring, not classic far-offshore, is what 80 percent of COOs plan to increase.
The solo-operator claim failed verification. "One person with AI does what a 10-person shop did" traces entirely to a cluster of content-farm sites repeating the same unsourced round numbers. Not one named 1-to-3-person CRM shop with verifiable evidence was found. The claim may still be true; the evidence for it does not currently exist in public.

Question 4Pricing models: what is winning, what is collapsing

Fixed fee per milestonewinning
The dominant model for defined implementation scope among named HubSpot partners: SmartBug $2,775 to $12,000 packages, Weidert $3,390 to $4,200 per Hub, IMPACT $3,000 to $25,000 plus. Hourly survives only for exceptions and scope changes. Medium-high
Retainer after go-livewinning
The second half of the standard two-stage deal: implement fixed, then run monthly. $1,000 to $5,000 a month support tier, $5,000 to $15,000 full service. Retainer adoption across agencies rose from 64 percent (2023) to 78 percent (2025), and retainer clients churn at 18 percent against 42 for project-only shops. High
Subscription, productizedgrowing minority
RevPartners is the strong named example: tiered productized implementation on a monthly subscription. Explicitly marketed as the differentiator by RevOps-native shops; a second identical example was not found. Low-medium
Pure hourlycollapsing as primary
Demoted to an exception mechanism, but not dead: survey data from 1,452 agency leaders (Promethean) shows most agencies run a stable blend, and 68 percent still quote $175 to $249 an hour as a live number. "AI killed hourly billing" is the narrative; a blended retreat is the data. Medium
Outcome or value baseddiscussed, rare
About 10 percent of agency revenue. Growing in talk far faster than in revenue. The WPP case is instructive: the CFO claims 20 to 25 percent of sales performance-linked; five months later the CEO names exactly one client on the model. Medium
Per AI agentforming next door
General AI-automation agencies quote discovery $1,500 to $3,000, build $5,000 to $25,000, then $300 to $1,500 a month per deployed agent. No named CRM implementation partner publishes a per-agent rate card yet; the model is bleeding in from the adjacent market. Low

Rate benchmarks, for calibration

HubSpot: independent consultants $100 to $150 an hour, certified US agencies $150 to $300, SmartBug specifically $185; typical implementation projects $12K to $60K, median near $15K. Salesforce: $100 to $200 an hour, mid-market projects $40K to $175K, annual retainers $36K to $120K. Senior RevOps practitioners: $200 to $400 an hour, with a $10K to $15K monthly retainer framed as "one senior RevOps hire, roughly". Zoho implementations price an order of magnitude lower and the sourced data is India-market. Medium

The missing case study, stated plainly. No named HubSpot, Salesforce or Zoho implementation agency was found saying "we charged X per hour, AI compressed delivery, now we charge Y" with dated numbers. The sharpest repricing evidence comes from ad agencies and generic AI shops, not CRM specialists. Whoever publishes that case first owns the conversation.

Question 5How agencies find and win clients now

The 2022 playbook was rank on Google, collect the vendor directory referral, sponsor the mega-conference. Every leg of that stool got shorter.

Vendor directoriesweakened
Still real, but the sold-versus-sourced split means a matched lead earns a third of the commission-years of an originated one. Vendors are telling agencies, structurally, to bring their own demand.
Content and SEOrebuilt around citation
Informational traffic down 30 to 60 percent on affected queries; commercial-intent traffic largely intact. The new game is being the source the model cites, which follows authority and third-party mentions, not rank.
Outboundre-professionalized
Signal-based Clay outbound replaced list blasting, and Clay competency itself became a marketable credential. OneAway: 190 plus Clay tables across 50 plus engagements.
Communities and owned eventsgrowing
Vetted RevOps communities (Wizards of Ops, RevGenius) surface partners by reputation. RevPartners built its own conference, Southbound, rather than only sponsoring INBOUND (now UNBOUND).
Referralsstill the largest
More load-bearing than ever as the other channels erode. The engineered version: a proprietary free tool as the referral trigger. Six & Flow attributed 25 percent of retainer business to free-tool wedges, though that figure dates to 2019. Low on currency, high on the pattern.

Question 6Business models: where the money actually is

Software resale is a side dish. Roughly 60 percent of a first-year CRM budget goes to services against 40 percent to licenses; implementation runs 1.5 to 3 times the annual license fee. The vendor commission (HubSpot's 20 percent for up to 3 years) is real but secondary, and the cost of staying eligible for it rose sharply in 2026. High

Margins: prescriptive targets say 50 to 60 percent delivery gross margin and 20 to 30 percent net; the measured all-agency average is nearer 13 percent, and implementation shops nearer 11. Eight-figure agencies in one benchmark average 25 to 32 percent net. The populations differ; do not mix the numbers. Agencies that narrowed their service offering grew 13 percent on average with 30 percent net margins. Medium

Valuations: 70 percent plus recurring revenue commands a premium of one to two multiple turns. Quoted multiples run from 2 to 5 times EBITDA for small owner-run shops to 9 to 12 times for PE-grade roll-up targets; any single multiple quoted without the deal-size band is suspect. Medium

The existential question, argued properly

AI eats the agency

  • Forrester predicts 15 percent of agency jobs cut in 2026.
  • HubSpot markets Breeze as needing no configuration or code, and its own playbook tells partners 45 percent of consultant activity is automatable.
  • 53 percent of agency owners call AI a serious threat, up from 44.
  • A client demanded an 80 percent fee cut claiming ChatGPT matched the agency's output. Low
  • Both vendors raised the bar to stay a partner at all: fewer partners survive.

The role shifts, the channel stays

  • 70 percent of Agentforce implementations still go through partners.
  • Salesforce is targeting $1B in partner incentives and doubled its partner fund, in the same period.
  • Complex rollouts still take months even with agentic tooling.
  • Consultant demand grew 10 percent globally, 72 percent in the US, in 2025.
  • The M&A wave is buying managed-services capability, not retreating from it.

The reconciliation the evidence supports: AI is compressing the low end of the work while both vendors consolidate the channel toward fewer, more specialized, more AI-fluent partners. Margin concentrates; the category does not disappear. The losers are the median generalist shops, and they are the majority.

Question 8The service catalogue: added, kept, retired

Growing
GTM engineering (the Clay stack), AI enablement and readiness (HubSpot's playbook prices it $50K SMB to $200K plus mid-market), fractional RevOps ($3K to $15K a month against $130K plus for the hire), agent governance (the layer the 40-percent-cancellation finding creates), custom development and warehouse integration (Snowflake, BigQuery syncs as a named line), CPQ (still resists templating).
Commoditizing
Basic implementation and standard config: HubSpot's own partner playbook says discontinue it, and HubSpot's mandatory $3K to $7K vendor onboarding now competes with partner onboarding directly. Simple migration: Lightfield shipped a one-hour AI migration agent in March 2026; complex dedup and custom objects still pay. Standard reporting: folded into maintenance tiers.
Absorbed by product
Data enrichment: HubSpot bought Clearbit (November 2023), now Breeze Intelligence firing automatically at form fill. Agent setup: shipped self-serve and usage-priced in the Spring 2026 Spotlight. What used to be an agency line item is now a checkbox.
The uneven front
Aptitude 8, RevPartners, Carabiner and elefante have visibly rebuilt their service pages around AI, as-a-service and warehouse work. SmartBug's live 2026 services page still reads like 2023, with zero AI language. The shift is real and it is not universal, which means the repositioning window is still open.

Question 9The best of the best, with the evidence

Twenty-five firms were enumerated; these are the ones whose claims survive sourcing, and what each one specifically proves.

Huble

HubSpot · 175 plus staff, 7 countries

HubSpot Global Partner of the Year 2024, ISO 27001 and 9001 certified. Proves: compliance certification is a real differentiator upmarket; nobody else in the cohort leads with it.

Aptitude 8

HubSpot · Atlanta, 71 staff

The only partner with a HubSpot Ventures investment; number two Global Partner of the Year 2024; acquired WORQFLOW for RevOps depth. Positions on custom development ("not possible in HubSpot" work) and an onshore-only team. Proves: technical depth plus vendor intimacy beats headcount; 71 people outrank far larger shops.

RevPartners

HubSpot + Clay · acquired June 2026

The only firm holding Elite status from HubSpot and Clay simultaneously; productized subscription pricing; runs its own conference (Southbound); acquired by Walker Sands to anchor a RevOps practice. Proves: the dual-credential, productized, own-audience model is what acquirers pay for.

SmartBug Media

HubSpot · 300 plus staff, PE-backed

Three-time North America Partner of the Year, 1,200 plus certifications, grew by acquisition (Globalia, 2023). Yet its services page still reads like 2023. Proves: scale and awards do not equal repositioning; the biggest partner is arguably the slowest mover.

Coastal Cloud and Cyntexa

Salesforce SMB

Coastal Cloud: 8,000 plus projects, ISG's number one Salesforce consultancy four years running, SMB implementations at $35K to $75K. Cyntexa: Summit tier December 2024, 900 plus projects in 55 countries, sells industry accelerators. Proves: at Salesforce SMB scale, fixed-price packages and prebuilt accelerators are the winning shape.

Winning by Design and Go Nimbly

RevOps consultancies

Winning by Design: trusted by one in eight public SaaS companies, published the recurring-revenue textbook, owns SPICED and the bowtie. Go Nimbly: fractional RevOps at $20K to $100K a month for Intercom, Twilio, Zendesk, Vanta. Proves: published IP and named logos are what let a services firm price like that.

Zenatta and PD Experts

Zoho · Pipedrive

Zenatta: Zoho Americas Partner of the Year 2024, 1,000 plus implementations. PD Experts: Pipedrive Global Partner of the Year two years running. Proves: the smaller ecosystems have their own leaders, and they win on volume and focus rather than IP.

The M&A ledger, 2023 to 2026: SmartBug plus Globalia (Oct 2023), Aptitude 8 plus WORQFLOW (Feb 2024), SBI Growth plus Carabiner (May 2024, the RevOps-as-a-Service pioneer), Walker Sands plus RevPartners (June 2026, 250 plus combined staff). Salesforce itself acquired Traction on Demand. PE money (American Discovery Capital, Mountaingate) is behind the roll-ups. High

Question 10Innovation, separated from repackaging

Every candidate judged on mechanism, not marketing.

MCP, headless CRM
Genuine A standardized, typed, agent-discoverable tool protocol is a real new primitive. Both top vendors shipped it in 2026.
Conversation-native CRM
Genuine Day.ai inverts the write path: the record is generated from conversation, not typed in. Built from proven parts, but the inversion is real.
Outcome and per-agent pricing
Genuine for the CRM category: Intercom, Zendesk, HubSpot and Salesforce all moved concrete pricing lines off the seat. Imported from usage-priced SaaS, not invented.
Agentic CRM
Incremental Tool-calling agents with write-back are a real upgrade over chatbots, and "agent washing" is an analyst-named epidemic. Gartner: 40 percent plus of projects canceled by 2027.
Self-implementing CRM
Incremental Attio's schema-flexible model with AI assist is genuinely better setup; "no admin needed" oversells what is still manual modeling.
AI SDRs
Split 11x: repackaging with documented fraud. Artisan: real product, retired its own slogan. Clay: genuine, because it claimed less.
Vertical AI CRMs
Repackaging Vertical CRM predates AI by decades; the AI layer is real but is not a new CRM primitive.
"The CRM disappears"
Repackaging The same pitch Streak and Copper made in the mid-2010s, re-argued with LLM auto-capture. Its one concrete implementation is Day.ai, graded above.

Question 11Eighteen frameworks, side by side

Every framework found, in one uniform shape: publisher, date, link, what it is for, and a verdict on whether it is new thinking. The folds carry stages, limits and adoption evidence, so any two can be compared field by field.

Verdicts:New thinkingIncrementalRepackagingEnduring canon

Clay GTM Engineering Framework

New thinking
Clay (Mishti Sharma, Varun Anand) · April 21, 2026 · clay.com/blog/gtm-engineering

How to build AI-driven revenue systems on top of the CRM instead of stitching point tools around it.

Stages, best use, limits, adoption.

Four layers: Data Foundation, Data Modeling, Data Activation, Agents; three laws: uniqueness, speed of iteration, engineering mindset. Best for: keeping AI outbound and enrichment workflows alive as tactics commoditize. Limits: vendor content justifying Clay's own product; assumes a GTM engineer role and budget. Adoption: distilled from playbooks deployed in 50 plus B2B SaaS companies. The only framework in the set built for a capability that did not exist before 2024.

Dynamics 365 Success by Design, AI-era update

New thinking
Microsoft FastTrack · ~2018, AI layer August 2026 · learn.microsoft.com

A governance layer over any project methodology, now with agentic design principles and a four-level AI maturity scale tied to project phases.

Stages, best use, limits, adoption.

Five phases: Discover, Initiate, Implement, Prepare, Operate, run through blueprint and go-live readiness reviews across 30 plus success measures. The 2026 AI scale: Human first, Copilot assisted, Humans with agents, Agent first with oversight. Best for: scoping how much autonomy an agent gets at each phase, the question no 2020 framework asked. Limits: Dynamics-specific; not a project methodology by itself. Adoption: the delivery backbone of Microsoft's own FastTrack program.

Microsoft Agentic AI Adoption Maturity Model

New thinking
Microsoft · 2026 · learn.microsoft.com/agents

Where an organization sits in adopting AI agents, from experimentation to agent-first with human oversight.

Stages, best use, limits, adoption.

Assessed across: strategy, process transformation, governance, value realization, architecture, operations, readiness, responsible AI. Best for: a shared vocabulary for how much autonomy to grant an agent now. Limits: brand new, tightly coupled to Copilot Studio, no long-run validation. Adoption: referenced inside Success by Design as its AI lens.

Winning by Design SPICED

Incremental
Winning by Design · ~2021 · winningbydesign.com

Situation, Pain, Impact, Critical Event, Decision: a customer-centric qualification language replacing BANT and MEDDIC.

Stages, best use, limits, adoption.

Best for: one qualification language across sales, marketing, CS and RevOps, configured into the CRM. Limits: needs retraining off BANT muscle memory; underdelivers without the bowtie beside it. Adoption: trained across 25,000 plus professionals. Genuinely better for subscription businesses; still the same five-letter-acronym genre it replaces.

Winning by Design Bowtie

Incremental
Winning by Design · 2019, revised October 2023 · winningbydesign.com

The customer journey extended past the close: onboarding, adoption, expansion, as a data model the CRM can carry.

Stages, best use, limits, adoption.

Shape: two triangles joined at the close: lead generation, lead development, sale, then onboarding, adoption, expansion. Best for: giving post-sale motions defined stages and metrics inside the CRM, which the funnel never had. Limits: built for recurring revenue; awkward for one-time purchases. Adoption: operationalized as a reference data model by RevOps tooling vendors.

Salesforce Well-Architected

Enduring canon
Salesforce Architects · 2023, revised 2024 onward · architect.salesforce.com

Trusted, Easy, Adaptable: the decision framework for Salesforce solutions that stay healthy as they scale.

Stages, best use, limits, adoption.

Best for: stopping technical debt before build, as the standard partner-architect review lens. Limits: Salesforce-only, principle-level, assumes an architect exists. Adoption: distributed through Trailhead and the Dreamforce architect track. Competent systematization, not new thinking.

Forrester B2B Revenue Waterfall

Enduring canon
Forrester, ex SiriusDecisions · 2002, relaunched 2021 · forrester.com

Tracks buying groups, not individual leads, through demand into revenue, split by acquisition, cross-sell, upsell and retention.

Stages, best use, limits, adoption.

Best for: auditable stage definitions where deals involve buying committees. Limits: needs buying-group data most CRMs do not capture natively; heavy to instrument. Adoption: the core content track of Forrester's B2B Summits and the stage-mapping standard for enterprise RevOps teams.

Prosci ADKAR, applied to CRM adoption

Enduring canon
Prosci (Jeff Hiatt) · 1998, guidance current · prosci.com

Awareness, Desire, Knowledge, Ability, Reinforcement: diagnoses exactly where CRM user adoption breaks per person.

Stages, best use, limits, adoption.

Best for: separating "knows the CRM, does not want it" from "wants it, was never trained", instead of one undifferentiated adoption problem. Limits: individual-level; needs an organizational frame beside it. Adoption: built from 700 plus change projects; Prosci's research across 2,600 practitioners shows well-managed change meets objectives at nearly twice the rate.

IDIC

Enduring canon
Peppers & Rogers · 1993, formalized 1999, 3rd ed. 2016 · wiley.com

Identify, Differentiate, Interact, Customize: the original blueprint most CRM strategy still traces back to.

Stages, best use, limits, adoption.

Best for: forcing a CRM strategy to start from customer segmentation and value, not from what the software can do. Limits: predates modern CRM software and AI-scale personalization. Adoption: still the canonical CRM strategy framework in business-school and consulting literature three decades on.

Gartner AI Maturity Model

Incremental
Gartner · ~2019, agentic coverage 2025 to 2026 · gartner.com (gated)

Five levels across seven pillars, from awareness to transformational; locates why a pilot is not scaling.

Stages, best use, limits, adoption.

Pillars: strategy, portfolio, governance, engineering, data, operating models, people. Best for: finding the actual bottleneck, which is usually governance or operating model, not the model. Limits: subscription-gated; the finding that most enterprises are stuck at experimentation is sobering, not actionable. Adoption: Gartner's 2025 survey: 45 percent of high-maturity organizations keep AI projects running 3 plus years, against 20 percent of low-maturity ones. The same CMMI shape maturity models have used since the 1990s.

McKinsey AI Trust Maturity Model

Incremental
McKinsey · 2025, agentic update 2026 · mckinsey.com

Governance maturity, scored 0 to 4 across five dimensions, now including agentic controls; separate from technical capability.

Stages, best use, limits, adoption.

Best for: surfacing the gap between "we use AI in the CRM" and "we can govern what the agent is doing". Limits: the global average score is 2.0 of 4, so it benchmarks a problem more than it solves one. Adoption: survey-based across a global enterprise panel.

Gartner Revenue Operations Model

Incremental
Gartner · ~2021 · gartner.com

Three maturity stages for coordinating strategy, process, data and technology across marketing, sales and service.

Stages, best use, limits, adoption.

Best for: a readiness diagnostic before committing to a reorg or re-platform; enter at any function. Limits: deliberately avoids fixed milestones, so it is not a build plan. Adoption: Gartner forecast 75 percent of the highest-growth companies running RevOps by 2026, from under 30 percent.

HubSpot Flywheel

Repackaging
HubSpot (Brian Halligan) · September 2018 · hubspot.com/flywheel

Attract, Engage, Delight in a circle, customer at the center; retention as the growth engine.

Stages, best use, limits, adoption.

Best for: reframing CRM investment around retention and referral. Limits: critics correctly note it relabels loops the funnel always implied and gives no stage math. Adoption: the default mental model across HubSpot's platform, Academy and partner messaging since 2018. The emphasis shift is valuable; the mechanism is not new.

RevOps Maturity Model, the category convention

Repackaging
RevPartners and at least five others · ~2022 onward · blog.revpartners.io

Five stages scored on people, process, data and technology; republished under many agency brands with cosmetic differences.

Stages, best use, limits, adoption.

Best for: locating where a prospect actually is before proposing a rebuild, so scope matches readiness. Limits: the identical shape ships under Varicent, Hyperscayle and others, so it signals category membership more than proprietary insight. Adoption: its very ubiquity is the evidence; it is now the default way to sell a RevOps engagement. Publishing one is table stakes, not differentiation.

The reference shelf: four vendor systematizations

Enduring canon
Salesforce Data Governance (~2023) · Salesforce Customer 360 Guides (2019 on) · DAMA-DMBOK data quality (2009, rev 2024) · HubSpot RevOps Certification (~2021)

Four competent codifications worth using and not worth imitating: governance pillars, industry reference architectures, the seven data-quality dimensions, and a certifiable RevOps vocabulary.

The four, with links.

  • Salesforce Data Governance Framework: quality, security, architecture, stewardship. salesforce.com/data/governance
  • Salesforce Customer 360 Guides: industry blueprints and reference architectures, uneven depth by industry. salesforce.com/customer-360
  • DAMA-DMBOK: the vendor-neutral data-quality vocabulary: completeness, accuracy, consistency, uniqueness, validity, clarity, currency. dama.org
  • HubSpot RevOps Framework and Certification: people, process, technology, data; the ecosystem's shared baseline vocabulary and hiring credential. academy.hubspot.com

What this could mean for Avero

Avero is a GTM-engineering agency and HubSpot partner running delegated RevOps execution with a knowledge-base doctrine and a stated bet on operating clients' AI processes. Nothing below is a plan; these are the six questions the outside evidence puts on the table, for steering.

1. The origination question.

Both vendors just repriced waiting for referrals. The winning firms originate demand: own audiences, own tools, own conferences, Clay-based outbound. Which origination motion is Avero's to own?

2. The ladder question.

The three-phase story (contractor, operations manager, strategic architect) says margin is migrating to AI-ready architecture and data foundations. Avero's knowledge-retention and architecture-decisions value props already point up the ladder. What would make that the thing clients name unprompted?

3. The pricing question.

Fixed fee per milestone plus a run retainer is the evidenced winning shape, and nobody in CRM has published the "AI compressed our delivery, here is our new pricing" case study yet. Is that a case study Avero could be first to publish?

4. The governance question.

If 40 percent of agent projects die from missing risk controls, agent governance (what it may touch, when it escalates, how success is measured) is the highest-margin new line, and it matches the AIOps bet already voiced inside Avero. Who sells that in Israel today? The research found nobody.

5. The credential question.

Clay fluency became a co-equal credential next to HubSpot certification, and the dual-Elite firm just got acquired. Does the Avero roadmap include a second credential, and which one?

6. The data-layer question.

MCP makes the CRM a governed tool surface for agents, which is Avero's existing HubSpot-as-source-of-truth doctrine arriving as an industry. What would an MCP-era service catalogue line look like, priced?

What this research could not establish

No CRM-specific AI repricing case. The before-and-after evidence for AI breaking hourly billing all comes from ad agencies and generic AI shops, never a named CRM implementation firm.
No measured delivery-compression baseline. Faster onboarding is claimed universally and quantified against 2023 nowhere.
No verifiable solo micro-agency. The one-person-does-ten claims trace to content farms citing each other.
Zoho and Attio partner economics are gated. No public commission figures exist; anything quoted elsewhere is unverified.
Israel-specific agency data is absent. Every benchmark on this page is US, UK or India sourced. The local market comparison is a v2 candidate.
Search budget note. The research fleet exhausted its shared web-search quota mid-run; the pricing, services and work-style passes finished on targeted fetches. Named gaps above are where that bit.
The brief, verbatim. The eleven research questions below are Amit's own words, kept exactly. Operational instructions from the same brief (sub-agent design, output mechanics, internal context pointers) live in the vault record, not on this public page.

The eleven questions.

Keep the search VERY WIDE, then go DEEP wherever it gets interesting. Do not narrow early. 1. What is happening in the world of CRM right now? 2. How is the SOFTWARE evolving? Agentic CRM, CRM as a data layer, AI-native challengers, what the incumbents (Salesforce, HubSpot, Microsoft, Zoho, Pipedrive, Attio, Clay, Folk and whoever else matters) are actually shipping versus announcing. 3. How are the IMPLEMENTATION AGENCIES evolving? What does the work itself look like now versus three years ago, and what is disappearing. 4. PRICING MODELS. Hourly, fixed, retainer, subscription, outcome, per-seat, per-agent, value-based, productised. What is winning and what is collapsing. 5. GTM MODELS. How these agencies find and win clients now. 6. BUSINESS MODELS. Partner/reseller economics, IP and product-ised service, recurring versus project revenue, margin structure. 7. WORK STYLE. Team shape, delivery method, how AI changed who does what, offshore versus onshore, pods, async. 8. SERVICES. What is in the catalogue now, what is being added, what is being retired. 9. WHAT ARE THE BEST OF THE BEST DOING? Name real firms and say what specifically makes them the best. Evidence, not adjectives. 10. WHAT INNOVATION are we seeing? The genuinely new things, separated from repackaging. 11. FRAMEWORKS. Are there new published frameworks for this kind of work and for making it succeed? Bring in every framework you find, each organised the same way, with publisher, date and a link. Present them so I can compare them side by side.