Numberai Inc v. Flai Tech Inc

I. Executive Summary and Procedural Information

• Parties & Counsel:
  • Plaintiff: NumberAI, Inc. d/b/a Numa (Delaware)
  • Defendant: Flai Technologies Inc. (Delaware)
  • Plaintiff's Counsel: Morris, Nichols, Arsht & Tunnell LLP; KXT Law, LLP
• Case Identification: 1:26-cv-00894, D. Del., 07/22/2026
• Venue Allegations: Venue is asserted in the District of Delaware on the basis that Defendant is a Delaware corporation and therefore resides in the district.
• Core Dispute: Plaintiff alleges that Defendant's AI-powered customer communication platform for automotive dealerships infringes two patents related to automated, communication-based intelligence engines.
• Technical Context: The technology at issue involves using artificial intelligence and machine learning models to analyze customer communications, automate responses, and manage business information and workflows in real-time.
• Key Procedural History: The complaint notes that U.S. Patent No. 11,553,055 is a divisional of U.S. Patent No. 10,917,483, and that both patents share the same specification and figures. This relationship suggests that claim construction arguments and prosecution history from one patent may be relevant to interpreting the other.

Case Timeline

Date	Event
2017-06-22	Earliest Priority Date for '483 Patent and '055 Patent
2021-02-09	U.S. Patent No. 10,917,483 Issues
2023-01-10	U.S. Patent No. 11,553,055 Issues
2026-07-22	Complaint Filed

II. Technology and Patent(s)-in-Suit Analysis

U.S. Patent No. 10,917,483 - Automated Communication-based Intelligence Engine

Issued February 9, 2021 ('483 Patent)

The Invention Explained

• Problem Addressed: The patent describes a business environment where managing high volumes of customer communications across fragmented channels is challenging Compl. ¶¶23-24 Prior systems often relied on manual review, static routing rules, or primitive analytics, which were inefficient, prone to error, and could not dynamically maintain an accurate, up-to-date context for each communication session '483 Patent, col. 2:5-67
• The Patented Solution: The invention proposes a system that uses historical communication data, including successful and unsuccessful outcomes (i.e., "conversions"), to train a "likelihood-of-conversion model" '483 Patent, col. 16:9-14 During an active communication session, this model analyzes the ongoing conversation to determine the probability of a conversion. This calculated likelihood is then used as a control value to decide whether to continue handling the session with an artificial response system or to hand it off to a human agent, thereby optimizing resource allocation '483 Patent, col. 17:40-47 '483 Patent, FIG. 9
• Technical Importance: This approach claims to improve the functioning of computerized communication systems by replacing static routing with a dynamic, state-driven process that conserves processing and network resources by allocating human intervention to sessions with the highest potential value Compl. ¶13 Compl. ¶16

Key Claims at a Glance

• The complaint asserts independent claim 1 Compl. ¶50
• Essential elements of Claim 1 include:
  • Receiving a plurality of communications from a plurality of secondary entities during a plurality of electronic communication sessions, wherein the plurality of communications include a plurality of conversions comprising a plurality of sale transactions.
  • Training a likelihood of conversion model based on the plurality of communications and the plurality of conversions.
  • Receiving via a network particular communications from a particular secondary entity during a particular communication session.
  • Using an artificial response system to respond to the particular communications.
  • Analyzing the particular communications to determine a particular likelihood of conversion based on the likelihood of conversion model, the particular likelihood comprising a likelihood that the particular secondary entity will engage in a particular sale transaction.
  • Handing off the particular communication session to a particular user from the artificial response system based on the particular likelihood of conversion.
• The complaint states infringement of "one or more claims... including without limitation claim 1," reserving the right to assert other claims Compl. ¶50

U.S. Patent No. 11,553,055 - Automated Communication-based Intelligence Engine

Issued January 10, 2023 ('055 Patent)

The Invention Explained

• Problem Addressed: The patent identifies the difficulty of maintaining accurate and synchronized information (such as business hours, FAQs, or service availability) across multiple, separate network platforms '055 Patent, col. 2:7-16 This often leads to conflicting or outdated information being presented to customers and requires redundant manual processes to correct Compl. ¶15
• The Patented Solution: The invention describes a system that first monitors various communications to build and maintain an "entity model" containing business-specific data '055 Patent, col. 11:46-51 The system then monitors a particular communication, specifically an "online chat message," and applies the entity model to it to generate an update '055 Patent, col. 11:57-61 Crucially, this generated update is then applied via a network to automatically change electronically published information on a separate platform, such as an FAQ or hours-of-operation listing '055 Patent, col. 11:62-12:13 '055 Patent, FIG. 8
• Technical Importance: This process claims to improve networked computer platforms by creating a closed-loop system that uses live communications to automatically keep published business information accurate, thereby reducing redundant processing, conserving network bandwidth, and improving the reliability of the platform's electronic outputs Compl. ¶14 Compl. ¶16

Key Claims at a Glance

• The complaint asserts independent claim 1 Compl. ¶57
• Essential elements of Claim 1 include:
  • Monitoring a plurality of communications of a particular entity.
  • Building an entity model of the particular entity based on the communications, the model comprising entity data.
  • Monitoring at least one particular communication comprising an online chat message between the entity and a user.
  • Applying the entity model to the online chat message to generate an update to the entity data.
  • Applying the update to a particular platform by changing at least one of electronically published FAQs, an hours-of-operation listing, or an event listing via an application program interface enabled by the platform.
• The complaint reserves the right to assert other claims beyond claim 1 Compl. ¶57

III. The Accused Instrumentality

Product Identification

• The "Accused Flai Product" is an AI-powered customer communications and customer experience platform for automotive dealerships, offered under the name "Flai" Compl. ¶37

Functionality and Market Context

• The Flai platform is designed to handle customer communications for automotive dealerships across multiple channels, including calls, texts, online forms, and chats Compl. ¶38
• Its alleged functionality includes an "AI receptionist" that answers questions and routes callers, and a "sales-focused AI" that captures leads, answers questions, and responds based on buyer intent Compl. ¶39 It also provides service scheduling and appointment-booking by checking a dealership's live schedule Compl. ¶40
• The complaint alleges the product integrates with existing dealership technology stacks, including schedulers, Dealer Management Systems (DMS), and Customer Relationship Management (CRM) systems Compl. ¶41 Compl. ¶43
• It is alleged that the Accused Flai Product builds, maintains, and applies dealership-specific information and AI models based on dealership communications, workflows, and operational data to generate responses, routing decisions, and updates to dealership data Compl. ¶¶44-45

No probative visual evidence provided in complaint.

IV. Analysis of Infringement Allegations

The complaint references exemplary claim charts in Exhibits 3 and 4, which were not filed with the complaint Compl. ¶50 Compl. ¶57 The analysis below is based on the narrative infringement allegations provided in the body of the complaint.

'483 Patent Infringement Allegations

Claim Element (from Independent Claim 1)	Alleged Infringing Functionality	Complaint Citation	Patent Citation
receiving multiple communications ... including communications that result in conversions and sales-related transactions	The Flai product allegedly receives communications from dealership customers that result in conversions such as lead captures, test-drive bookings, and vehicle purchases.	¶51(a)	col. 15:58-16:8
training and/or updating a likelihood-of-conversion model based on the received communications and conversion-related outcomes	The Flai product allegedly trains or updates a model based on outcomes like lead qualification and appointment bookings.	¶51(b)	col. 16:9-14
receiving, via a network, particular communications from particular dealership customers...	The Flai product allegedly receives communications from specific customers during particular sessions.	¶51(c)	col. 17:27-31
using Flai's artificial response system... to respond to those particular communications	The Flai product's AI receptionist and sales AI functionality allegedly respond to communications by answering questions and collecting details.	¶51(d)	col. 17:36-39
analyzing the particular communications to determine a particular likelihood of conversion based on the likelihood-of-conversion model	The Flai product allegedly analyzes communications to determine the likelihood a customer will engage in a sales-related transaction.	¶51(e)	col. 17:40-43
handing off the particular communication session from Flai's artificial response system to a dealership user... based on the determined likelihood of conversion	The Flai product allegedly hands off sessions to a human user based on the determined likelihood of conversion, buyer intent, or other factors.	¶51(f)	col. 17:44-47

• Identified Points of Contention:
  • Scope Question: Claim 1 requires training a model based on a "plurality of sale transactions." A point of contention may be whether the alleged training data, which includes "lead qualification" and "appointment bookings" (Compl. ¶51(b)), meets the "sale transactions" limitation, or if that term requires a completed vehicle purchase.
  • Technical Question: A key question will be whether the accused system's routing decisions are governed by a "likelihood-of-conversion model" as specifically claimed, or by a more general "intent" detection mechanism (Compl. ¶51(f)). The evidence will need to show not just a handoff, but a handoff based on a specific type of predictive model trained on past conversion outcomes.

'055 Patent Infringement Allegations

Claim Element (from Independent Claim 1)	Alleged Infringing Functionality	Complaint Citation	Patent Citation
monitoring multiple communications of a particular dealership or dealership entity	The Flai product allegedly monitors communications including calls, texts, messages, and online inquiries.	¶58(a)	col. 11:41-45
building an entity model... comprising dealership-specific entity data such as workflows, departments, customer and lead information	The Flai product allegedly builds an entity model with dealership-specific data like workflows, departments, and customer information.	¶58(b)	col. 11:46-51
monitoring at least one particular communication... comprising an online chat message	The Flai product allegedly monitors specific communications including texts and chats.	¶58(c); ¶58(f)	col. 11:52-56
applying the entity model to the online chat message... to generate an update to the entity data of the entity model	The Flai product allegedly applies its entity model to generate updates for lead details, customer needs, and appointment information.	¶58(d); ¶58(g)	col. 11:57-61
applying the update to the particular platform by changing at least one of electronically published FAQs, an hours-of-operation listing, or an event listing	The Flai product allegedly applies updates to dealership systems like DMS and CRM through networked integrations, such as changing a service appointment in a scheduler.	¶58(e); ¶58(h)	col. 11:62-12:13

• Identified Points of Contention:
  • Technical Question: A central issue will be whether the accused product's integrations with dealership platforms (Compl. ¶58(h)) perform the specific function of automatically changing a listing, as required by the claim. The analysis will need to distinguish between simply logging or transmitting data to a system (e.g., a DMS) for human review versus programmatically altering a stateful listing, such as an available appointment slot in a scheduler.
  • Scope Question: The claim recites changing specific types of "electronically published" information: "FAQs, an hours-of-operation listing, or an event listing." It will be a point of dispute whether an update to an internal dealership system like a CRM or DMS (Compl. ¶58(e)) falls within the scope of this limitation, or if the term requires a more public-facing publication.

V. Key Claim Terms for Construction

'483 Patent, Claim 1: "likelihood of conversion model"

• Context and Importance: This term is the core of the '483 Patent's inventive concept. The definition will determine whether a general-purpose AI routing engine infringes, or if the claim is limited to a specific type of predictive model trained on particular outcomes. Practitioners may focus on whether the defendant's system performs the specific "training" and "analyzing" functions associated with this model.
• Intrinsic Evidence for a Broader Interpretation: The specification describes the model in terms of probability, stating it "performs a test to score a particular message or conversation" and that the "score represents the probability that the message or conversation will lead to a conversion" '483 Patent, col. 16:15-19 This could support a construction covering any probabilistic scoring system for routing.
• Intrinsic Evidence for a Narrower Interpretation: Claim 1 explicitly requires the model to be trained on a "plurality of sale transactions." This language could be used to argue for a narrower construction where the model must be trained specifically on completed sales, not just on intermediate steps like lead captures or inquiries.

'055 Patent, Claim 1: "applying the update to the particular platform by changing at least one of electronically published FAQs, an hours-of-operation listing, or an event listing"

• Context and Importance: This limitation defines the ultimate action of the claimed process. The dispute will likely turn on what it means to "change" a "listing" on a "platform." It distinguishes the invention from systems that merely collect and display information. Practitioners may focus on the required nexus between the "online chat message" and the automated "changing" of the separate platform's data.
• Intrinsic Evidence for a Broader Interpretation: The specification states the system is used "to update other services and products that the entity relies on" '055 Patent, abstract, which could be argued to encompass a wide range of internal and external systems.
• Intrinsic Evidence for a Narrower Interpretation: The claim language itself provides a specific, exemplary list ("FAQs, an hours-of-operation listing, or an event listing"). This may be used to argue the "platform" must be one that publishes such information. Further, the specification notes that the system "programmatically registers changes to incorrect listings" and directs an interface to "register the change with the Google™ server," suggesting a direct, automated modification of a data record on a third-party or external system '055 Patent, col. 12:5-6 '055 Patent, col. 12:13-16

VI. Other Allegations

The complaint does not provide sufficient detail for analysis of indirect or willful infringement. The infringement counts are pleaded exclusively under 35 U.S.C. § 271(a) for direct infringement Compl. ¶50 Compl. ¶57

VII. Analyst's Conclusion: Key Questions for the Case

1. A central question for the '483 Patent will be one of technical and definitional scope: Does the accused AI platform's use of "buyer intent" to route communications constitute a "likelihood-of-conversion model" trained on "sale transactions" as required by the claim, or is there a fundamental mismatch in how the predictive model is trained and applied?

2. For the '055 Patent, a key evidentiary question will be one of technical operation: Does the accused product's integration with dealership platforms perform the specific, two-part function of (1) generating an update from an "online chat message" and (2) using that update to automatically "change" a separate, electronically published listing (like a scheduler), or does it merely log or forward information for subsequent human action?

3. A recurring issue for both patents will be one of claim construction: How broadly will terms like "sale transaction" ('483 Patent) and "electronically published... listing" ('055 Patent) be defined? The answers will determine whether internal dealership workflows and data (e.g., lead captures, CRM entries) fall within the scope of claims that appear directed at more definitive commercial outcomes and public-facing information.