The buzzed-about however nonetheless stealthy New York Metropolis startup Augmented Intelligence Inc (AUI), which seeks to transcend the favored "transformer" structure utilized by most of at the moment's LLMs corresponding to ChatGPT and Gemini, has raised $20 million in a bridge SAFE spherical at a $750 million valuation cap, bringing its complete funding to just about $60 million, VentureBeat can solely reveal.
The spherical, accomplished in below every week, comes amid heightened curiosity in deterministic conversational AI and precedes a bigger elevate now in superior levels.
AUI depends on a fusion of the transformer tech and a more recent know-how referred to as "neuro-symbolic AI," described in better element beneath.
"We understand that you may mix the brilliance of LLMs in linguistic capabilities with the ensures of symbolic AI," mentioned Ohad Elhelo, AUI co-founder and CEO in a current interview with VentureBeat. Elhelo launched the corporate in 2017 alongside co-founder and Chief Product Officer Ori Cohen.
The brand new financing consists of participation from eGateway Ventures, New Period Capital Companions, current shareholders, and different strategic buyers. It follows a $10 million elevate in September 2024 at a $350 million valuation cap, coinciding with the firm’s introduced go-to-market partnership with Google in October 2024. Early buyers embody Vertex Prescription drugs founder Joshua Boger, UKG Chairman Aron Ain, and former IBM President Jim Whitehurst.
In keeping with the corporate, the bridge spherical is a precursor to a considerably bigger elevate already in superior levels.
AUI is the corporate behind Apollo-1, a brand new basis mannequin constructed for task-oriented dialog, which it describes because the "financial half" of conversational AI — distinct from the open-ended dialog dealt with by LLMs like ChatGPT and Gemini.
The agency argues that current LLMs lack the determinism, coverage enforcement, and operational certainty required by enterprises, particularly in regulated sectors.
Chris Varelas, co-founder of Redwood Capital and an advisor to AUI, mentioned in a press launch supplied to VentureBeat: “I’ve seen a few of at the moment’s high AI leaders stroll away with their heads spinning after interacting with Apollo-1.”
A Distinctive Neuro-Symbolic Structure
Apollo-1’s core innovation is its neuro-symbolic structure, which separates linguistic fluency from job reasoning. As an alternative of utilizing the most typical know-how underpinning most LLMs and conversational AI techniques at the moment — the vaunted transformer structure described within the seminal 2017 Google paper "Consideration Is All You Want" — AUI's system integrates two layers:
-
Neural modules, powered by LLMs, deal with notion: encoding consumer inputs and producing pure language responses.
-
A symbolic reasoning engine, developed over a number of years, interprets structured job parts corresponding to intents, entities, and parameters. This symbolic state engine determines the suitable subsequent actions utilizing deterministic logic.
This hybrid structure permits Apollo-1 to keep up state continuity, implement organizational insurance policies, and reliably set off software or API calls — capabilities that transformer-only brokers lack.
Elhelo mentioned this design emerged from a multi-year information assortment effort: “We constructed a shopper service and recorded tens of millions of human-agent interactions throughout 60,000 stay brokers. From that, we abstracted a symbolic language that defines the construction of task-based dialogs, separate from their domain-specific content material.”
Nevertheless, enterprises which have already constructed techniques constructed round transformer LLMs needn't fear. AUI desires to make adopting its new know-how simply as simple.
"Apollo-1 deploys like every fashionable basis mannequin," Elhelo instructed VentureBeat in a textual content final night time. "It doesn’t require devoted or proprietary clusters to run. It operates throughout commonplace cloud and hybrid environments, leveraging each GPUs and CPUs, and is considerably extra cost-efficient to deploy than frontier reasoning fashions. Apollo-1 may also be deployed throughout all main clouds in a separated setting for elevated safety."
Generalization and Area Flexibility
Apollo-1 is described as a basis mannequin for task-oriented dialog, that means it’s domain-agnostic and generalizable throughout verticals like healthcare, journey, insurance coverage, and retail.
In contrast to consulting-heavy AI platforms that require constructing bespoke logic per shopper, Apollo-1 permits enterprises to outline behaviors and instruments inside a shared symbolic language. This method helps sooner onboarding and reduces long-term upkeep. In keeping with the staff, an enterprise can launch a working agent in below a day.
Crucially, procedural guidelines are encoded on the symbolic layer — not realized from examples. This allows deterministic execution for delicate or regulated duties.
As an illustration, a system can block cancellation of a Fundamental Financial system flight not by guessing intent however by making use of hard-coded logic to a symbolic illustration of the reserving class.
As Elhelo defined to VentureBeat, LLMs are "not a great mechanism whenever you’re searching for certainty. It’s higher if you already know what you’re going to ship [to an AI model] and all the time ship it, and you already know, all the time, what’s going to return again [to the user] and how one can deal with that.”
Availability and Developer Entry
Apollo-1 is already in energetic use inside Fortune 500 enterprises in a closed beta, and a broader common availability launch is predicted earlier than the top of 2025, based on a earlier report by The Data, which broke the preliminary information on the startup.
Enterprises can combine with Apollo-1 both through:
-
A developer playground, the place enterprise customers and technical groups collectively configure insurance policies, guidelines, and behaviors; or
-
A normal API, utilizing OpenAI-compatible codecs.
The mannequin helps coverage enforcement, rule-based customization, and steering through guardrails. Symbolic guidelines permit companies to dictate mounted behaviors, whereas LLM modules deal with open-text interpretation and consumer interplay.
Enterprise Match: When Reliability Beats Fluency
Whereas LLMs have superior general-purpose dialog and creativity, they continue to be probabilistic — a barrier to enterprise deployment in finance, healthcare, and customer support.
Apollo-1 targets this hole by providing a system the place coverage adherence and deterministic job completion are first-class design objectives.
Elhelo places it plainly: “In case your use case is task-oriented dialog, it’s a must to use us, even if you’re ChatGPT.”
