Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Salesforce expanded partnerships with OpenAI and Anthropic to embed their large language models into Agentforce 360.

At Dreamforce 2025, Salesforce said it has expanded a partnership with OpenAI so customers can access Agentforce 360 in ChatGPT. Salesforce customers will be able to access sales records, review conversations and build Tableau visualizations from within ChatGPT.

• Salesforce makes its Agentforce 360 case to be your AI agent platform
• Salesforce's acquisition of Apromore highlights how process intelligence, agentic AI converging

For Salesforce, which is betting on Slack as an agentic OS and front end to Agentforce, the OpenAI partnership appears to be a nice hedge to play into whatever the new enterprise user interface exists. For OpenAI, the Salesforce partnership lands a big enterprise software vendor for its application strategy. At OpenAI Dev Day, CEO Sam Altman walked through APIs and a strategy that brings various apps to ChatGPT. Most of the applications demonstrated were decidedly consumer.

In a statement, Salesforce said its customers will be able to use GPT-5 to build AI agents and prompts within the Salesforce Platform. For Salesforce, the move highlights how its platform can be headless. Salesforce also sees its partnership with OpenAI as a way to enable its retail customers to sell more.

Salesforce noted that the partnership with OpenAI gives it a big footprint with ChatGPT's 800 million weekly users and Slack's 5.2 billion weekly messages. The goal: Embed Salesforce in a "preferred surface environment."

Altman said the Salesforce partnership is "an important step in how AI can improve daily workflows under our efforts together." Salesforce CEO Marc Benioff said the partnership brings the instant recommendations or insights consumers expect to enterprises.

Integrations include:

• Agentforce 360 apps in ChatGPT.
• Instant Checkout and Agentforce Commerce, via Agentic Commerce Protocol, will be integrated. Salesforce is supporting Agentic Commerce Protocol via a partnership with Stripe.
• ChatGPT in Slack, which will bolster search results and insights.
• Codex, an OpenAI agent for coding, will be available in Slack.
• OpenAI's models will be in Agentforce 360.
• OpenAI will join a host of providers such as Anthropic in Slack Marketplace.

The Anthropic partnership is focused on regulated industries. Under an expanded partnership, Anthropic's Claude will be available in Agentforce for regulated industries that need tighter data control.

In addition, Claude will be integrated into Slack for enterprise workflows. Claude will analyze documents, access internal data and enable decisions without app switching.

Salesforce is providing the trust layer and customers such as CrowdStrike and RBC Wealth Management are using Claude through Amazon Bedrock in Agentforce.

Key points about the Salesforce-Anthropic partnership:

• Claude can be a preferred AI model provider for regulated industries.
• Anthropic and Salesforce will work on industry specific offerings for regulated industries.
• Anthropic is the first LLM vendor to be fully integrated within Salesforce's trust layer. All of Claude's traffic is contained in Salesforce's virtual private cloud.
• Anthropic Claude for Financial Services with Agentforce Financial Services and AI agents can work across platforms.

For Salesforce, the Anthropic offering meets enterprises where they are. Anthropic is the enterprise choice in many cases and has a strategy that more aligned with enterprise software. By adding Anthropic to its trust layer, Salesforce is giving the LLM provider more enterprise scale while offering tight integration that benefits both companies.

• Anthropic, Microsoft, OpenAI build out AI agent use cases
• Anthropic expands international presence, enterprise reach
• Anthropic follows enterprise software industry playbook, hires Smith as commercial chief

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Nvidia's DGX Spark mini supercomputer is now available for $3,999 from Nvidia and PC makers such as Acer, ASUS, Dell Technologies, GIGABYTE, HPI, Lenovo and MSI and it'll be interesting to see how impacts inference and physical AI plans.

DGX Spark’s sales will kick off Wednesday at Nvidia and partners. DGX Spark launched in March after being introduced as Project Digits at CES 2025. The system, basically a supercomputer in a mini-desktop form factor, is aimed at developers. The idea was that developers would use DGX Spark to develop and run models locally and then upload them to the cloud for production.

The reality is there will be plenty of DGX Spark buyers who want to say they have a supercomputer even though they won't be useful for everyday tasks.

For Nvidia, the launch of DGX Spark is a way to highlight the power of its stack. DGX Spark is built on Nvidia Grace Blackwell architecture and integrates the company's GPUs, CPUs, networking and software libraries.

In other words, DGX Spark may be handy for inference at the edge and physical AI. DGX Spark has a petaflop of AI performance, 128GB of unified memory and the ability to run inference with up to 200 billion parameters.

For all the hubbub of Nvidia CEO Jensen Huang delivering the first DGX Spark system to Space-X's Elon Musk, the real impact will likely be at the edge. A few thoughts:

• DGX Spark has more power than Nvidia's 2016 DGX-1 system from 2016 with a much lower price and power profile.
• The system has Nvidia's AI software stack preinstalled, which will only solidify the company's dominance in AI applications.
• Nvidia said early DGX Spark customers are validating and optimizing tools for the system.
• The real impact of DGX Spark will come from enterprise deployments at the edge. Can DGX Spark be hardened for the manufacturing floor, supply chain and other less hospitable places? Most likely.
• Robotics use cases were cited by Nvidia in a blog post.
• Ultimately, DGX Spark's impact is likely to be seen in physical AI.

For now, it'll be interesting just to see how sales go. DGX Spark systems will be available for sale Oct. 15.

• SoftBank's ABB robotics purchase boosts physical AI plans
• Physical AI, world foundation models will move to forefront
• Nvidia launches Cosmos models, aims to expand physical AI, industrial reach
• Humanoid robots near inflection point courtesy of AI

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Amazon Web Services said Amazon Bedrock AgentCore is generally available and said the AgentCore software development kit has seen more than 1 million downloads from customers.

AgentCore is the latest entry into the agentic AI platform race. Google Cloud launched Gemini Enterprise, which absorbs what used to be Agentspace. Salesforce outlined Agentforce 360, the fourth version of Agentforce in 12 months. Those agentic AI platforms landed in recent days, but there are multiple vendors such as ServiceNow, Boomi and UiPath all looking to be the management layer for AI agents.

What AWS is hoping to do with AgentCore, initially outlined at AWS Summit New York, is provide a horizontal stack that can enable customers to deploy, operate and secure AI agents with all the necessary building blocks such as tools, memory, data and workflows. AWS considers AgentCore to be a key platform for operationalizing AI agents.

• AWS launches Amazon Quick Suite, aims to automate business workflows
• The AWS AI Strategy: Playing the Long Game Infrastructure-Style
• AWS' AI strategy: Jassy's long talking and the big picture
• AWS Q2 sales growth 17.5%, nears $124 billion annual revenue run rate
• AWS launches Bedrock Agent Core, custom Nova models
• AWS' practical AI agent pitch to CxOs: Fundamentals, ROI matter

By operating horizontally, AWS sees AI agents as a natural extension to its Bedrock platform, which provides model choices.

In a recent briefing, AWS noted that AgentCore is designed to give enterprises the flexibility to start with any model or agent framework. Early customers include Cox Automotive, Druva, Cohere Health, Ericsson, Sony and Thomson Reuters to name a few. AWS has also surrounded AgentCore with big partners such as Accenture, Deloitte and Salesforce.

Key points about AgentCore include:

• AgentCore includes composable services that can be used together or independently.
• The platform supports multiple AI agent frameworks such as CrewAI, Google ADK, LangGraph, LlamaIndex, OpenAI Agents SDK and Strands Agents and models on Amazon Bedrok, OpenAI or Gemini.
• AgentCore includes AgentCore Code Interpreter, AgentCore Browser and AgentCore Gateway, which takes existing APIs and AWS Lambda functions and makes them agent friendly tools with connections to third-party data and Model Context Protocol servers.
• AWS has also added AgentCore Identity, which provides authentication and authorization for AI agents.
• Dashboards and tracking is provided via AgentCore Observability, which integrates with various monitoring tools.
• AgentCore Runtime automatically scales agent workloads and security so each agent session has its own isolated computing environment. AgentCore supports virtual private cloud deployments.

According to AWS, the goal of AgentCore is to get from pilot to production faster in a way that is secure with little friction for developers.

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

OpenAI will design its own AI accelerators and chips in deal with Broadcom that will start in the second half of 2026 and run through 2029.

According to the companies, the deal covers 10 gigawatts of custom AI accelerators. OpenAI has forged partnerships with AMD and Nvidia as it builds out its infrastructure.

• OpenAI, AMD ink big GPU deal: What it means for the rest of us
• Stargate ramps as OpenAI, Oracle, Softbank outline 5 new data centers
• Nvidia to invest $100 billion in OpenAI, which will deploy Nvidia next-gen AI infrastructure
• Watercooler debate: Are we in an AI bubble?

OpenAI and Broadcom will co-develop systems that include custom accelerators and Ethernet networking. These racks will be built using Broadcom components. The rationale behind the partnership is that OpenAI can better optimize its own stack. OpenAI is following the hyperscale cloud playbook for AI that includes big doses of Nvidia as well as custom-made GPUs such as Google's TPUs and AWS' Trainium.

The news isn't a big surprise given that Broadcom noted the strength of its pipeline on its most recent earnings call. It was widely assumed that OpenAI was the key customer.

OpenAI CEO Sam Altman said "developing our own accelerators adds to the broader ecosystem of partners all building the capacity required to push the frontier of AI."

Broadcom said that the OpenAI racks will include its portfolio of Ethernet, PCIe and optical connectivity as well as custom accelerators.

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Salesforce announced a series of product updates, a process intelligence acquisition and an Agentforce 360 architecture designed to make the case that it should orchestrate first and third party AI agents.

At Dreamforce 2025, Salesforce looked to differentiate itself from what's becoming a crowded agentic AI platform space. The cloud giants are looking to go horizontal with stacks to build, deploy and optimize AI agents. SaaS vendors are expanding into managing AI agents beyond their verticals. There's a set of neutral platforms such as Boomi and UiPath. And everyone is trying to layer process intelligence into the agentic AI stack via tuck-in acquisitions.

• Salesforce's acquisition of Apromore highlights how process intelligence, agentic AI converging
• Salesforce heads into Dreamforce 2025: A look at the big themes
• Is Salesforce ripe for Serious Disruption? And if so, from Whom?
• Salesforce raises prices across multiple products including Slack
• Salesforce revamps Agentforce pricing with Flex Credits: What you need to know

If Dreamforce 2024 was that coming out proof-of-concept party for Agentforce, Dreamforce 2025 is about production and scale with Agentforce 360. The challenge for customers is piecing together the barrage of news announcements and product updates and sorting through the keynote flash to understand the big picture. The Agentforce pacing is brisk with four iterations in 12 months.

"What we want to see is that all companies become agentic enterprises," said Parker Harris, co-founder of Salesforce and CTO of Slack. "We're entering a world where it's going to be humans and AI working together. It's the biggest transition in technology I've ever experienced."

Harris, who sets the technical vision for Salesforce, would know. He helped start the SaaS transition and has been through the mobile and social shifts.

• Enterprise AI: It's all about the proprietary data
• The big AI, SaaS, transformation themes to watch in 2025’s home stretch
• AI agents, automation, process mining starting to converge
• LLM giants need to build apps, ecosystems to go with the models
• Pondering the future of enterprise software

Here are some of the big Salesforce themes to note.

Agentforce 360 expands Salesforce's reach from CRM

Salesforce announced a unified agentic AI stack that is now called Agentforce 360, which includes Agentforce Platform, Data 360, Customer 360 Apps and Slack. The general idea is to connect humans, AI agents, apps and data across the entire enterprise.

Srini Tallapragada, President and Chief Engineering and Customer Success Officer, said the goal with Agentforce is to ensure that AI isn't disconnected from context. "Agentforce 360 is a deeply unified platform," he said. "We don't want customers stuck in what I call the pilot bucket. What we had to solve is to ensure that AI is not disconnected from context."

If this approach sounds familiar, it's because multiple vendors are trying it. Salesforce and ServiceNow appear to be on a collision course. You don't have to look farther than ServiceNow's foray into CRM and Agentforce IT Service to realize the two vendors are rivals. The main thing to note is that the Salesforce and ServiceNow have plenty of joint customers and CRM and ITSM both remain fragmented enough for both to grow while talking smack.

Agentforce 360 includes the following:

• Hybrid Reasoning Engine. Within Atlas, Salesforce is providing a deterministic control system called Agent Script. The goal here is to provide predictable agent behavior while keeping the creativity. The Hybrid Reasoning Engine will be in beta in November and provide customers the ability to build via natural language or code.
• A revamped Agentforce Builder, which will be in beta in November. Agentforce Builder will provide predictable logic, a conversational build and test loop and a natural language approach.
• Agentforce Studio and Agent Observability close the loop to create what Salesforce hopes is a continuous agent improvement flywheel.
• Voice integration. Agentforce 360 includes seamless voice capabilities to hand off calls between AI agents and humans.
• Context engineering with the ability to process unstructured data as well as structured.
• Multi-agent orchestration. MuleSoft becomes the Agent Fabric that can govern and coordinate multiple agents including those from third parties.
• Apps for multiple verticals and use cases.
• Informatica will fit into this mix on the data front when Salesforce's acquisition closes. Informatica's role in the Agentforce stack: Turn enterprise data into a searchable catalog as Data Cloud activates data, MuleSoft connects any system and Tableau is an insights engine.

Of those news items, it's worth noting that Salesforce is trying to thread the needle between predictability, which is an enterprise must-have, and creativity, a customer must have. By controlling AI agent behavior, Salesforce is trying to give enterprises the ability to provide a unified brand voice and customer experience as AI agents and humans run a CX relay race.

Perhaps the biggest takeaway is that Salesforce has noted that its platform can be headless across channels and run in the background.

Holger Mueller, an analyst at Constellation Research, said:

"While it's competitors are working on single products and second versions, Salesforce has a suite of AI offerings available for customers with it's typical 360 brand. All of this is another data point on how Salesforce has a lead of one or two years over its key competitors."

Real returns and value

Salesforce highlighted a series of customers deploying Agentforce, provided returns and walked through the production process.

Yes, Salesforce spent a good amount of time detailing its success as its first customer. Salesforce internally handles 4.8 million customer conversations across multiple channels.

Salesforce also has a year's worth of deployments under its belt. The company has developed blueprints for enterprises deploying AI agents to move to production faster. Salesforce said it has more than 200 agentic industry workflows across 15 industries such as automotive, communications, consumer goods, manufacturing, public sector and retail.

• For AI agents to work, focus on business outcomes, ROI not technology
• BT150 zeitgeist: Consumption-based pricing angst abounds

CIOs will be more interested in hearing from the parade of customers. DirecTV saved 300,000 hours through Agentforce, Under Armour doubled case deflection rates, Williams Sonoma uses AI agents for personalized interior design and Heathrow Airport deployed an AI agent to improve customer experiences.

Salesforce also noted that Agentforce can work across multiple verticals including manufacturing where Eaton saw a 71% reduction in cost per service call.

Salesforce also noted that it has landed enterprises that are standardizing on Agentforce. For instance, Salesforce said Vonage is using Agentforce for its platform as is Statista and Reddit.

Here are a few customer vignettes from a Dreamforce briefing call:

• Indeed's Lisa West, Vice President of Automation at the job listing site, said the company went with Agentforce because it was heavily invested in the Salesforce platform. "We have so much rich context inside of the Salesforce ecosystem so it was quite natural to add agents to that," said West.
• Peter Burns, Director of Marketing, Digital and E-commerce at Heathrow Airport, used Agentforce to create Hallie, a personal customer agent. Heathrow uses Salesforce's platform for its customer contact center. Burns said Heathrow uses Agentforce to provide a simple interface to a complicated technical ecosystem. "We think about creating experiences in the front and back of house. In the back of house, Agentforce and Salesforce power our customer center. Agentforce helps answer and manage those conversations," said Burns. "Hallie was launched as a customer-facing agent that answers questions and manages more of your bookings. We have more than 80 million passengers and as much as we want to individually hold their hands through Heathrow, it's not practically possible. Hallie and Agentforce allow us to effectively scale personalized customer experience."
• Sameer Hassan, Chief Technology and Digital Officer at Williams Sonoma, said the retailer is homing in on using Agentforce to create personalized agents focused on interior decorating and design. Hassan said AI agents were initially seen as a productivity play, but evolved into tools that could drive growth. "Yes, we're seeing the efficiency gains, and yes, we are automating the repetitive task, but AI is actually amplifying that creative and strategic work that our people are doing," said Hassan. "When you equip brilliant minds with this kind of technology, we are actually seeing better strategic and creative outcomes from folks whether it's product development, supply chain, merchandising, photography and visualization or technology. It's also about amplifying your people."

Slack as an "agentic OS"

One of Salesforce's unique approaches to Agentforce and AI agents is that it is betting Slack can be the front end to all Salesforce applications.

"Slack is not only becoming the front-end of Salesforce and Work 360, but it's becoming your agentic OS where you can search, collaborate and act across all the people in your organization, data and connect with agents," said Parker. "We're reimagining Salesforce and Slack. Maybe you don't log in to Salesforce or even see it, but it's there. It's coming to you in Slack because that's where you're getting your work done with native AI and enterprise AI search. Then you can unify your entire enterprise in one agentic OS and bring in other AI agents."

This strategic bet on Slack comes as OpenAI is betting ChatGPT can be the front end of all applications, ServiceNow is using its AIx approach with a similar conversational UI and Workday bought Sana to do something similar.

Add it up and there's an industry-wide bet that the ChatGPT interface and multi-model LLM approach will be the entry point into all applications.

If Slack winds up being that agentic OS to enterprise AI, Salesforce will be differentiated. Perhaps OpenAI will be that OS for consumers and Salesforce will happily take the enterprise AI agent OS turf.

Parker said that Slack will connect easily with OpenAI, Google and multiple partners via Model Context Protocol (MCP). Slack AgentExchange will be available in the fourth quarter with Slack personal agent landing in the first quarter and agentic search will be in pilot in January.

Slack will get:

• Slack-first Apps such as Agentforce sales, service, ITSM and HR.
• An agent with always-on expertise.
• Enterprise search.
• A reimagined Slackbot that’s context aware.
• Multiple integrations via MCP.

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Neptune Insurance Holdings has successfully completed its initial public offering with a business model that revolves around applying data science, machine learning and artificial intelligence to provide flood insurance. The company may be well on its way to being an AI exponential.

To the uninitiated, flood insurance is a bit of a disaster in the US. According to FEMA, flooding is the most common and costly natural disaster with damages topping $40 billion annually according to the Congressional Budget Office. Flood maps are outdated, many building owners in flood zones forgo insurance and most policies are provided by the National Flood Insurance Program (NFIP).

In high-risk states such as Florida, Texas and Louisiana residential flood insurance penetration is below 13% combined. The federal insurance typically has high premiums and low coverage limits that fail to address rebuilding costs. NFIP has a cover limit of $250,000 compared to Neptune's $7 million.

Insurers stay away from flood insurance due to the risk and potential costs. Neptune Insurance is betting that it can apply AI to manage risk, provide a good customer experience and insure more homes with more affordable coverage.

In its SEC filing, Neptune Insurance, which has 60 employees, said: "We believe the NFIP’s legacy pricing model, cumbersome processes, and limited coverage have created significant market dislocation and inefficiencies, resulting in a compelling opportunity for private flood insurers like Neptune to capture market share."

Why would Neptune want to play ball in the flood insurance market? AI, machine learning and data science. Neptune may be among the first of the AI-native industry-focused enterprises to go public, but it certainly won't be the last.

• The road to AI Exponential will be bumpy
• AI Exponentials Will Rule the World
• Radically Rethink Old Models To Enable the Power of AI
• Enterprise AI: It's all about the proprietary data
• Get Insights Newsletter

Neptune's AI stack, which runs on the cloud but appears to be mostly custom built (or a very well-kept secret), enables the company to carry out underwriting risk selection and pricing, aggregation and carrier assignment in less than two seconds via its Triton underwriting engine. Neptune has no human underwriters.

The Triton underwriting engine has a lifetime written loss ratio of 24.7% from 2016 inception through June 30. From 2018 to 2024, NFIP has a written loss ratio of 86% according to FEMA. The broader property and casualty industry has a written loss ratio of 54%, according to NAIC data.

Key points about Neptune's underwriting approach:

• The company leverages its own claims data as well as industry wide signals. "With over a quarter billion dollars in paid claims across our portfolio since inception, and our extensive analysis of industry claims and performance data, we have developed deep insights into identifying and managing properties with the highest probability of flood losses," said the company.
• Neptune has proprietary models focused on the characteristics and factors leading to large-scale losses. Neptune has a unit called Neptune Data Science Group that creates the models and predictive analytics that fuel the company.
• Models are continually optimized.
• Pricing models are driven by models that account for flood risk data, customer behavior and market dynamics. The company doesn't use manual adjustments or broad risk groupings to price insurance. "Behavioral economics plays a crucial role in determining how customers perceive and value coverage, and incorporating customer behavioral factors into our methodology enables us to tailor pricing to increase adoption while maintaining underwriting integrity. By understanding not only the risk, but also the behavior of customers, we believe we can optimize premium structures to drive growth and retention," said Neptune in regulatory filings.
• Neptune disaggregates policies across geographic boundaries to manage risks. In addition, Neptune doesn't have balance sheet insurance risk or claims handling responsibility because it uses a network of 26 reinsurance providers.

The data stack

We reached out to Neptune for an interview, but didn't get a reply. It's unclear what cloud Neptune uses and checks with the hyperscalers resulted in no comment.

Mike Dezube, Chief Data Science Officer at Neptune, co-founded Charles River Data, which was a data science consulting firm acquired by Neptune in May 2024. Dezube worked at Google on search, machine learning and healthcare and has focused on AI, GIS data and decision engines.

CTO Brad Schulz has experience in insurance technology platforms, flood insurance and behavioral marketing.

While Neptune's stack appears to be build over buy it does leverage a few key providers focused on geospatial data and automated workflows.

• AI Forum Washington, DC 2025: Everything we learned

Here’s a quick look at Neptune’s data stack:

• Neptune Triton incorporates KatRisk APIs to overlay flood footprint data into exposure post natural disaster. The modeling tools bolstered Neptune's ability to manage risk.
• Neptune has used Ecopia AI for its building-based geocoding to price based on granular location.
• Neptune used ICEYE, a data provider on flood hazards.
• Customer experience and self-service operations are powered by Ada, which provides a front end to insurance buying. Ada's bots provide answers to customer questions about policy payments, endorsements and documents. Neptune uses an internal customer success team that uses Zendesk and Zoom to handle more complicated queries.

Neptune isn’t likely to turn up into a big enterprise software keynote, but there are lessons to be learned from its data and model as differentiator strategy.

Disruptor-like results

Neptune said it has surpassed 250,000 policies in force, but what stuck out about the company was its results and levels of automation.

The company's net profit per employee checks in at $750,000 and revenue per employee is $2.5 million from inception to date.

For the six months ended June 30, Neptune reported net income of $21.56 million on revenue of $71.42 million, up more than 32% from a year ago. In 2024, Neptune reported net income of $34.6 million on revenue of $119.3 million.

Those figures put Neptune into the AI native camp in Constellation Research's framework.

Now there are risks with Neptune's business. The company has weathered multiple hurricanes already, but natural disasters are always a risk. In addition, Neptune could take a hit from a fading housing market--especially in flood prone areas that boomed during the Covid pandemic.

But so far, Neptune appears to be a disruptor worth watching. The company priced its IPO Sept. 30 at $20 and hit a high of $33.23 Oct. 3 before settling into a range between $25 and $30.

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Salesforce said it has acquired Apromore, a process intelligence software provider, as it aims to bring process optimization to its Agentforce platform.

As previously noted, agentic AI will require a hefty dose of process mining, task mining and intelligence to really benefit enterprises. Without process knowhow, there's a risk you'll simply automate less than optimal processes. Agentic AI could mean you actually scale faulty processes.

Apromore, founded in 2009, had raised $30 million in funding and Salesforce is already an investor in the company. Other vendors have been touting process expertise with agentic AI platforms. ServiceNow in its latest platform release integrated process and task mining into its AI agent workflows. Microsoft previously acquired Minit, a process mining company. Workday talked process for financials and HCM. And UiPath is pure play process automation companies that has evolved to be an AI agent orchestration platform.

• See: AI agents, automation, process mining starting to converge
• Hi-Fi process management needs to be foundation for AI
• Time to Introduce Hi-Fi for Enterprise Processes

Terms of Salesforce's acquisition of Apromore weren't disclosed. The purchase, announced just before Salesforce's Dreamforce conference, is expected to close in the fourth quarter.

In a statement, Salesforce said Apromore will bring "deep domain expertise in process intelligence and optimization directly into the Salesforce platform." Salesforce added that Apromore will be able to provide a real-time view in how processes run across enterprise systems.

Specifically, Salesforce said Apromore will provide:

• Visibility across all business processes, systems and applications.
• A foundation to target optimal automation use cases using process intelligence. Agentic AI, like process mining, is a game of finding use cases, such as accounts payable, order-to-cash and procurement, which drive the most returns for early wins.
• Ongoing optimization using Apromore's process and task mining, digital twin and simulation, root cause analysis and compliance tools.
• Apromore has a system neutral no-code approach that already leverages MuleSoft and has connectors to multiple systems including SAP, ServiceNow, Oracle and others.

Here's a look at Apromore's software.

Apromore CEO Marcello La Rosa said joining Salesforce accelerates the company's plan to democratize process intelligence. He noted that "the majority of our customers already deploy our technology on Salesforce."

Steve Fisher, President and Chief Product Officer at Salesforce, said Apromore's integration into the company's platform will bring the ability "to unlock opportunities to measure, optimize, and automate through agentic process automation."

Constellation Research analyst Holger Mueller said:

"Salesforce keeps bolstering its PaaS layer for Agentforce and the acquisition of Apromore highlights the trend. Process intelligence graphs are a proven and successful approach to make sure agentic AI does not hallucinate and it's usually more reliable than data based RAG. Now the question is how fast it will be able to make process intelligence capabilities to show up in Agentforce."

More on process optimization:

• Process mining vendor Celonis sues SAP over data access
• Lowe's betting on AI to drive customer experience, optimize multiple processes
• Bloomfilter co-CEO Severinghaus on the intersection of process mining, software development
• UiPath: Process automation orchestration core of AI agent deployments
• Why enterprise, process workflows are the new battleground