An independent, funding-weighted map of the AI customer agent category as of Q4 2026. This edition replaces our Q2 2026 map, which is preserved as published, and grades the five outcomes that edition said it was tracking.

The structural change since Q2 is that this stopped being a startup category. Add up the acquisition dollars spent on AI customer agents in 2026 — Salesforce‘s $3.6B for Fin, NICE’s $955M for Cognigy, roughly $250M of enterprise value for LivePerson, up to $17M for Agency — and you get about $4.8B, which is roughly 1.8 times the combined lifetime funding of every AI-native startup named in this map. The money has moved from backing challengers to buying them.

The five outcomes we said we were tracking

1. Sierra‘s Q3 2026 ARR disclosure — did not happen. We wrote that “$250M+ supports the multiple” and “$200M flat triggers a structured down-round.” Sierra has disclosed nothing since reaching $100M ARR in November 2025, roughly seven quarters after launch. One undated 2026 comparison cites “$150M+”; we will not carry that as a figure. The absence is the finding: a company valued at $15.8B in May has not published a revenue number in eleven months.

2. Decagon‘s vertical-AI launch — not found. We said that without an e-commerce or prosumer-SaaS specialisation, “fast time-to-value stays a feature rather than a position.” No vertical launch has surfaced. The position remains a feature claim.

3. Foundation-model native enterprise agents — arriving, but not in the shape we described. We expected a native enterprise agent runtime that would collapse the wrapper layer. What shipped instead was distribution: Agent Skills moved toward an open standard with a partner directory, Claudeforce put Salesforce inside Claude with 37 prebuilt sales skills in August, and ZoomInfo is natively available in OpenAI Codex for Work. The pressure on wrappers is real and it is coming through the interface, not through a competing agent.

4. Salesforce M&A activity — fired, with a different target, and that is the most informative result on this list. We wrote that “a Sierra acquisition would be the cleanest exit signal in the category.” Salesforce did buy — it agreed to acquire Fin (formerly Intercom) for $3.6B on June 15, 2026, closing expected early 2027. Read the choice: offered a category at $15.8B or a 76%-resolution agent with 30,000 customers at $3.6B, the largest potential acquirer took the second. That reprices the whole “replace the human at enterprise scale” premium, because the most obvious buyer just revealed what it is willing to pay.

5. Voice-AI regulatory action — no action. Instead Bland raised a $50M Series C in June at a reported total past $100M and pointed it at regulated industries. Same conclusion as our AI SDR map: the trigger is still loaded, the direction arrived without it.

Funding-weighted view, Q4 2026

Vendor	Position	Disclosed total	Last round	Last disclosed valuation
Sierra	Replace (enterprise)	~$1.6B	$950M, May 2026	$15.8B
Decagon	Replace (mid-market)	~$481M	$250M, Jan 2026	$4.5B
Cresta	Augment	~$282M	$125M, Nov 2024	$1.6B (Mar 2022)
Netomi	Replace (mid-market)	$160M+	$110M, Apr 2026	Undisclosed
Bland	Voice	~$106M	$50M, Jun 2026	Undisclosed at Series C
Crescendo	Hybrid AI + human	$50M	Oct 2024	~$500M
Encore AI	Replace (mid-market)	$30M	$30M, Jul 2026	Undisclosed

Two corrections to the Q2 edition’s figures. Crescendo’s total financing is $50M, per the company’s own October 2024 release — not the ~$80M our Q2 map carried, a number that circulated widely in secondary coverage. And Cresta’s last disclosed valuation is $1.6B from March 2022, not the “~$1B post” we attached to its Series D; the Series D valuation was never disclosed. We also omitted Netomi from the Q2 map entirely, which was an error — it has raised $160M+ including a $110M Series C in April 2026.

Positions, re-scored

Replace the human, enterprise (Sierra) — held, but the exit price just got named

Sierra remains the position of record: $1.6B raised, $15.8B valuation, the Fortune-500 reference base. Nothing has gone wrong operationally that we can see. What has changed is external — the most likely acquirer in the category bought someone else for less than a quarter of Sierra’s mark, and Sierra has not published revenue in eleven months. Those two facts together are why we would not underwrite the multiple today, and they are both facts rather than opinions.

Replace the human, mid-market (Decagon, Netomi, Encore AI) — crowded

This is now the most populated position, and the one where the Q2 map’s analysis has aged worst, because we treated it as Decagon’s alone. Netomi’s $110M Series C and Encore AI’s $30M Series A both fund the same promise to the same buyer. Decagon’s reported median contract — near $400,000, single-source — is the only pricing anchor any of them has given, and a crowded position with one visible price is a position heading for discounting.

Augment the human (Cresta) — the quiet one

Cresta has not raised since November 2024 and its last disclosed valuation is four years old, yet it reportedly reached $100M ARR in 2026 (secondary). If that figure is close, Cresta is the least richly valued real revenue in this map. Our Q2 read — that regulated verticals and real-time assist are durable because replacement is legally constrained there — is the thesis we would still defend most confidently.

Hybrid AI + human (Crescendo) — thinnest

$50M total against competitors holding $481M and $1.6B. The hybrid thesis — buy a guaranteed outcome, let the vendor decide the AI-to-human mix — is sound and we have argued it is honest. It is also the position least able to survive a price war, and the incumbents described below are about to start one.

Platform- and incumbent-owned agents (new, and not a startup position)

The Q2 map treated incumbents as a threat. In Q4 they are a position, because they now own products: Fin inside Salesforce (pending close), Cognigy inside NICE ($955M, closed September 2025, operating independently including its partnerships with Genesys and Avaya), Pinkfish inside Genesys, and Agency inside Klaviyo. Add NICE’s CXone Mpower Agents (June 2025), the four-part agentic stack Genesys announced at Xperience 2026 on roughly $2.9B of cloud ARR, and Five9’s Parloa partnership from May 2026.

Our Q2 threat call was that “Genesys, NICE and Five9 will each ship native AI agents in 2026/27” and flood the budget line with default-on alternatives. That was the most accurate call in the Q2 map, and it landed early. We also wrote that Agentforce “will likely be good enough by 2028.” Salesforce evidently disagreed with its own timeline and spent $3.6B to skip to the answer.

What this means for the buyer

The 2026 advice was pick on buyer fit, not brand. That still holds, with one addition that did not apply six months ago: ask every AI-native vendor what happens to your contract if they are acquired, and get the answer in writing. Four vendors in or adjacent to this category changed owners in 2026. The change-of-control clause has stopped being boilerplate.

And if you already pay for a contact-center platform, run its native agent against your top three ticket categories before you evaluate anything else. Not because it will be better — it probably will not be — but because the pricing conversation with a best-of-breed vendor changes completely once you can credibly say the default-on option clears your bar.

Predictions for 2027, with resolution criteria

1. Sierra does not publish an ARR figure before June 30, 2027. Resolves: a company-stated number in a release or a major outlet. Eleven months of silence after a $950M round is a choice, not an oversight.

2. No AI-native vendor in this map is acquired above $3.6B by Q2 2027. Resolves: a dated announcement. Salesforce set the comp; buyers rarely pay more than the last printed number for a category they have already bought into.

3. At least one of Decagon, Netomi or Encore AI announces vertical specialisation by Q2 2027. Resolves: a product announcement naming a specific industry. A crowded horizontal position with one visible price forces differentiation.

4. Crescendo raises, is acquired, or exits the map by Q4 2027. Resolves: a dated announcement, or its removal from this map for lack of activity. $50M is not enough to hold a position against $4.8B of acquisition money.

Methodology: funding and M&A figures are as recorded on our funding tracker, which requires an exact date plus either a primary source or two independent sources; undisclosed deal values are recorded as undisclosed rather than estimated. The $4.8B acquisition total sums Salesforce–Fin ($3.6B announced), NICE–Cognigy ($955M), SoundHound–LivePerson (~$250M implied enterprise value) and Klaviyo–Agency (up to $17M per Klaviyo’s 10-Q). ARR and contract-value figures are company-stated or single-source press reports and are not independently verified; we label them where used. Position assignments reflect each vendor’s primary GTM message, not internal roadmaps. GTMLens accepts no vendor money, paid placements or affiliate commissions — see our editorial policy.