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

Atlassian said its latest AI features and Rovo, a generative AI assistant that operates across the company's platform, are generally available across the company's products. Atlassian also introduced Rovo Agents.

The company said it will offer Rovo for annual subscriptions at $20 per user per month, monthly at $24 per user a month and consumption pricing in mid-2025. Licensing plans will be based on Rovo use per site where any users with access to the site is a billable user. Enterprises would only pay once per billable user.

Enterprise software vendors have been tweaking monetization models as some vendors focus on consumption or even conversations with an AI agent.

In May, Atlassian launched Rovo with the following core components:

• Rovo Search, which will comb through content wherever it is stored (Google Drive, Microsoft SharePoint, GitHub, Slack etc.), and query across applications. Rovo Search will identify team players, projects and information needed to make decisions. Rovo Search will connect niche and custom apps via API and have enterprise-grade governance to data governance.
• Insights, which are delivered via knowledge cards that offer context about projects, goals and teammates.
• Rovo Chat, a conversational bot that is built on company data and learns as it goes.

In a briefing, Jamil Valliani, Head of Product AI at Atlassian, cited early customers who have boosted efficiency by about 25% using Rovo with their development teams. Rovo beta testers said they've saved 1 hour to 2 hours of time saved per week.

• 13 artificial intelligence takeaways from Constellation Research’s AI Forum

Atlassian is connecting Rovo Search via connectors and connecting that data to Rovo Chat and throughout the platform.

The company also outlined Rovo Agents, which operate out-of-the-ox from Atlassian's marketplace partners. Atlassian is providing more than 20 out-of-the-box agents and tools to build your own Rovo Agents with low and no code tools.

According to Atlassian, Rovo Agents can speed up the development process by automatically generating code plans, code recommendations and pull requests based on task descriptions, requirements and context.

Other updates for Atlassian Intelligence include:

• Jira Service Management will use AI to group related alerts and surface critical incidents, suggest right resources and subject matter experts. The AIOps capabilities also capture incident timelines, generate post-incident reviews and summarize details.
• Jira Service Management virtual service agent will automate support across multiple platforms and add new onboarding and automation enhancements.
• Loom will get AI-powered automated workflows via integrations with Jira and Confluence.

Atlassian's AI additions will be critical to the company's future growth. In August, Atlassian projected first quarter revenue of $1.149 billion to $1.157 billion, below the consensus estimate of $1.16 billion. For fiscal 2025, Atlassian projected revenue growth of about 16%, below the 18% expected by Wall Street.

• Constellation ShortListâ„¢ Incident Management
• Constellation ShortListâ„¢ DevOps

The company at the time cited uncertain macroeconomic conditions and an evolving go-to-market strategy.

Speaking at an investment conference, Atlassian Chief Operating Officer Anu Bharadwaj said early adoption of Atlassian Intelligence and Rovo has been strong.

"Thousands of customers have adopted Atlassian intelligence already so far, and I'm very pleased with the repeated usage that it gets because one of the interesting things about AI is where are the use cases where you can unlock tangible productivity benefits. I think it is still early innings, so I’m very much looking forward to seeing how that plays out."

Regarding pricing, Bharadwaj said Atlassian has raised prices for its cloud products over time as it has added AI, automation and new features. "The price increases are very much in tune with the amount of customer value that we are able to deliver," he said. "In terms of seat-based versus not, I do think that there is an interesting exploration there around consumption-based pricing, which we will really think through, especially in an AI world, where we talk about virtual agents, which will be different than a seat-based model."

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

GE Healthcare has been working on machine learning, deep learning and artificial intelligence for years, but now the company sees an inflection point where generative AI can transform healthcare from products to workflow to efficiencies that improve the customer experience.

Parminder Bhatia, Chief AI Officer of GE Healthcare, said the emergence of multimodal large language models (LLMs) can uniquely improve healthcare, which is built on everything from different modalities, imaging data, clinical notes, voice interaction, electronic health records and other data.

Before GE Healthcare, Bhatia oversaw generative AI and large language models at Amazon Web Services. His group worked on Amazon Q and Amazon Bedrock. GE Healthcare and AWS recently announced a partnership to transform healthcare with a focus on purpose-built generative AI models using services such as Bedrock.

We caught up with Bhatia, an AI 150 inductee, at Constellation Research's AI Forum in New York to talk shop. Here's a look at the takeaways.

GE Healthcare's approach to AI. Bhatia has been in his current role for about 18 months overseeing the strategy and vision for AI at GE Healthcare. For GE Healthcare, the AI strategy revolves around the AI going into the MRI, CT and X-ray machines as well as digital platforms that focus on clinical and operational efficiencies across a hospital.

"There's a lot of focus on how we build these technologies that can really streamline workflow," said Bhatia. For instance, AI in an MRI machine that can reduce scan time by 50% with the same quality doubles the efficiency and productivity of the workforce.

• 13 artificial intelligence takeaways from Constellation Research’s AI Forum
• BioNtech, InstaDeep bet on genAI models to advance R&D, drug discovery, cancer treatment
• AI 15O's Matt Lewis on GenAI adoption, psychology and life sciences
• Welldoc Chief AI Officer Anand Iyer on AI, data and healthcare transformation

Other examples of AI's role at GE Healthcare include AI in ultrasound equipment that can act as a copilot, remote scans and imaging and technologies that "improve the efficiencies and accelerate getting better diagnosis, solving problems in treatment and cancer areas as well," said Bhatia.

GE Healthcare has been a pioneer within machine learning and deep learning for more than a decade and has the highest number of FDA approved app authorizations three years in a row.

Why generative AI and healthcare go together. Bhatia said LLMs have been all the talk, but the excitement around them is that they are multimodal. That ability to be multimodal means they apply well to healthcare.

He said:

"These technologies are truly multimodal in nature and that means they're more tailored for healthcare, which consists of data coming from different modalities, imaging data, clinical notes, voice interaction, your EHRs and other data. As these technologies were being built out it made sense for me to get back into healthcare. It's the perfect opportunity to apply these applications."

Patient experience and AI. Bhatia said AI will ultimately have an impact on the patient experience as workflows and staffing levels are improved for diagnosis to screening to treatment and therapy. GE Healthcare Command Center is using AI to streamline hospital operations, manage staffing and send triggers for actions. While many of those technologies don't affect the patient directly, the patient experience is improved with capacity planning.

"These technologies streamline operations and that becomes relevant across a spectrum of things," said Bhatia. "Patient guidance will also be key as we take care from inside the hospital to outside with patient monitoring and virtual care at home."

These hospital workflows will give a longitudinal patient view across care that improves experiences, he said.

Indeed, GE Healthcare recently acquired MIM Software, a company that manages workflows from diagnosis to treatment and therapy. A few recent developments with MIM include:

• GE Healthcare's MIM Software received FDA approval to perform absorbed dose calculation of radionuclides using a Monte Carlo method that will improve dose planning with automation and standardization.
• GE Healthcare showcased AI-enhanced radiation therapy tools.
• The company is also using AI to enhance cardiac imaging and assessments.

Personalization of care. Bhatia said AI will also play a big role in personalized treatment for cancer that deliver targeted radiation to kill cells.

"In the next three to five years, you're going to have thousands of variations in which these different radiopharmaceutical drugs can be given to the individual patients," he said. "MIM Software is designed to address the complexities that happen across the system, where it provides solutions to navigate the expanding landscape of personalized treatment."

Bhatia added:

"A lot of these things are starting with operational efficiency, but also combining multimodal data. I think that's where AI is becoming a key enabler, not just at the diagnosis level, but health clinicians can streamline the longitudinal view of the patient's data, which is truly multimodal. That technology and data can really streamline the operations, which has impact on better therapy and more personalized therapy for patients as well."

GE Healthcare's approach to AI. Bhatia said the company is taking a hybrid approach to AI and investing in talent focused on cloud and AI. "We are bringing a lot of that muscle for cloud and AI across the spectrum," he said. "That becomes the key component as we're looking into a lot of problems and challenges as well."

The hybrid strategy will mean "a lot of things happen on prem and a lot of things will happen in the cloud to accelerate and transform," said Bhatia. With AWS, GE Healthcare will look to building its own foundational models as well as using multiple LLMs for everything from workflows to equipment to treatments and imaging. Bhatia said:

"The partnership we announced with AWS is about strategy and foundational model building for building our own proprietary genAI, streamlining workflows and developing use cases. The partnership is really 1+1 is greater than 2 because you get a lot of benefits from security and scale with AWS and GE Healthcare being in more than 160 countries."

This approach to hybrid AI will also mean multiple partnerships for clinical research. Ultimately, GE Healthcare wants to be able to predict if a patient is going to skip or arrive late to appointments adapt workflows, and build in flexibility, said Bhatia.'

Model choice. Bhatia said flexibility with foundational models is critical. "One model is not going to solve all problems and you'll have to look to the clinical side and the operational side of things," he said. "The first place AI can have an impact is to alleviate cognitive and data overload to highlight what's relevant."

Bhatia added that models will also need to be adapted and used for specific use cases. Open-source models also have potential to be adapted for a specific use cases.

AI as a horizontal and vertical tool. Bhatia said it's important for AI leaders to think about generative AI as a horizontal enabler and a technology that can be used to drill down in specific areas. He said:

"You can build these AI algorithms for breast cancer, but they can easily be adapted to prostate cancer or lung cancer. And I think that's where these technologies are becoming really game changer. How do you adapt them, not just looking into the vertical side of things, going from diagnosis to treatment therapy and the entire patient journey, but also how they can be adapted across the spectrum?"

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 launched a NIM Agent Blueprint for cybersecurity as it continues to expand use cases for its microservices and AI agent platform.

At its AI Summit in Washington DC, Nvidia outlined its NIM Agent Blueprint for container security. The cybersecurity NIM Agent Blueprint combines Nvidia's Morpheus cybersecurity AI framework, Nvidia cuVS and Rapids data analytics to accelerate vulnerabilities (CVEs) at scale.

The cybersecurity blueprint is included in Nvidia AI Enterprise, the GPU giant's flagship software platform for AI applications.

Nvidia has had a steady stream of NIM Agent Blueprint news as it aims to make agentic AI more commonplace in enterprises.

• Accenture to use Nvidia stack for agentic AI
• Nvidia drops NVLM 1.0 LLM family, revving open-source AI
• Nvidia highlights algorithmic research as it moves to FP4
• Nvidia launches NIM Agent Blueprints aims for more turnkey genAI use cases
• Nvidia shows H200 systems generally available, highlights Blackwell MLPerf results

According to Nvidia, its NIM Agent Blueprint for container security enables enterprises to use generative AI to digest information and then explain vulnerabilities using natural language. Companies can then create agents for cybersecurity workflows.

Nvidia added that Deloitte is among the first to use Nvidia NIM Agent Blueprint for container security in its cybersecurity applications.

Here's a look at the architecture.

Among other notable items from Nvidia at its AI Summit.

• The company outlined healthcare systems adopting digital health agents.
• Nvidia, Yale and Moderna released a paper on quantum machine learning and drug discovery.
• Nvidia outlined how the SETI Institute was using AI to search for radio signals in space.

More on agentic AI:

• Enterprises leading with AI plan next genAI, agentic AI phases
• With Salesforce push, AI Agents, agentic AI overload looms
• 13 artificial intelligence takeaways from Constellation Research’s AI Forum
• Boomi CEO Lucas: AI agents will outnumber your human employees soon

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

SAP at its TechEd conference delivered its share of AI agent headlines as its Joule generative AI becomes one assistant across its platform, but as a more practical matter more developer options for ABAP and data lake capabilities will have a much larger impact.

The goal for SAP is to move its custom SAP ECC code to S/4 clean code as soon as possible and the company outlined a series of moves to make that happen faster with a big assist from generative AI.

At TechEd, SAP said it will enable ABAP developers to generate rate high-quality code with its Joule generative AI copilot to comply with SAP's ABAP cloud development model. According to SAP, "Joule will also be able to generate explanations for legacy code, making it easier to modernize legacy codebases and migrate to a clean core." ABAP is a programming language that runs in the SAP ABAP runtime environment, created and used by SAP for the development of application programs.

Constellation Research analyst Holger Mueller did a deep dive on the implications for ABAP developers, which will get extended customer fields, business logic and processes. Mueller noted that the ABAP additions to SAP Build will give SAP the ability to update legacy code at scale. ABAP has 2 million active developers.

By the end of 2024, SAP Build will include access from ABAP development tools and environments for SAP S/4HANA Cloud. The integration will enable developers to create and monitor ABAP Cloud projects in SAP Build.

• SAP’s Joule everywhere plan: ‘We are not developing AI just for the sake of AI”
• SAP Q2 cloud ERP revenue up 33%, sees restructuring hitting 9,000 to 10,000 jobs
• Amazon Bedrock integrated into SAP AI Core, SAP to use AWS chips
• SAP acquires WalkMe for $1.5 billion
• Boomi aims to ease SAP Datasphere migrations

The other big move by SAP revolved around new embedded data lake features for SAP Datasphere. By the end of the fourth quarter, Datasphere will have a data lake option to complement existing storage. Businesses will be able to analyze data across hybrid environments and preserve context and logic.

SAP said the data lake capabilities include:

• An integrated object store for more efficient data transformation and processing.
• Spark compute based on existing Datasphere data integration.
• The ability to access data on integrated object stores without physically copying the data.
• Ultimately, users will be able to import and integrate data at scale from SAP sources and non-SAP sources.

Mueller said:

"SAP shows broad investment across its technology platform, which is the key focus for all enterprise software vendors, readying their SaaS suites for the era of Infinite Computing. SAP with its new data lake capabilities adds (for the first time) object storage abilities. With this move, SAP for the first time gives customers the chance to holistically build next generation of applications powered by AI on its platform. Equally is important that SAP finally shows some love to its 2M+ active ABAP developers with support of ABAP in SAP Build, as well as making Joule available in ABAP. Critical for the ecosystem is that SAP now finally allows partners to build custom ABAP Code in SAP S/4 HANA Cloud Public Edition – a key move to help SAP customers to upgrade to S/4HANA. The impact of these three makes the solid progress on AI pale, where SAP shows the right execution – widening and deepening Joule capabilities. What stands out for Joule is the design point to have a single AI assistant across SAP."

Here's a look at everything announced at TechEd:

• SAP Build will get the ability to give Joule custom skills as well as use SAP HANA Cloud to ground large language models using its vector engine.
• Joule will get multiple AI agents that will combine business function expertise with the ability to carry out complex workflows. Joule will bring together specialized AI agents in areas like supply chain, procurement and finance.
• SAP will add two out-of-the-box autonomous AI agent use cases including dispute management for incorrect or missing invoices, duplicate payments and the like and a financial accounting use case to streamline financial processes.
• Joule will support 80% of SAP's most used business tasks by the end of the year. Joule will also be available in SAP Service Cloud and SAP Concur as well as SAP S/4HANA Cloud Public Edition. Simply put, Joule will be integrated into all of SAP's clouds.
• SAP Knowledge Graph launched as a business context tool that is preloaded with ABAP tables, CDS views, APIs, and key data models so enterprises can ground AI models.
• The company said it added Anthropic Claude 3.5 Sonnet via Amazon Bedrock to its generative AI hub along with the addition of IBM Granite foundation models, Meta Llama 3.1 and Mistral Large 2 and Codestral.
• SAP Generative AI Hub gets improvements to customize pre-trained AI models, a new software development kit, and new regions from the big three hyperscale cloud providers.

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

One of Intuit's big bets for the years ahead is to disrupt the mid-market ERP business market with its Intuit Enterprise Suite as it aims to fill a gap between QuickBooks Online Advanced and costly ERP implementations that are required when companies grow.

The company launched the Intuit Enterprise Suite last month as it took its unified platform, powered by Intuit Assist, a common data store, services from its various offerings and generative AI, and took aim at an $89 billion total addressable market for larger, mid-market businesses.

Intuit launched Enterprise Suite in the US for multi-entity, service and project-based businesses. These early adopters are providing a feedback loop for Intuit to continue to iterate on its platform.

Sasan Goodarzi, CEO of Intuit, laid out the importance of Intuit Enterprise Suite during the company's recent Investor Day. "We have no intention to serve enterprise businesses, but every intention to serve large mid-market businesses" said Goodarzi, who said there's a massive gap in the ERP market between companies that are growing and complex but don't have the time for an ERP implementation. "You go out and talk to large customers and their words are 'big ERP is an organ transplant, it's too expensive and it's not about the yearly expense as much as it is the couple of years it taes to shift to an ERP solution.'"

Goodarzi explained that Enterprise Suite has been in the works for a long time, but AI has enabled it to serve customers far north of $3 million in revenue that have multiple locations, the need for segment reporting and various requirements. In other words, Intuit has a continuum that serves prosumers with QuickBooks, small businesses with QuickBooks Online Advanced and midmarket firms with Enterprise Suite.

• Intuit embraces LLM choice for multiple use cases
• Intuit sees 'green shoots' from its generative AI strategy
• Intuit’s bets on data, AI, AWS pay off ahead of generative AI transformation (PDF)

Among early adopters of Intuit Enterprise Suite, annual revenue per customer (ARPC) is about $20,000. That sum is a win for Intuit and businesses since Enterprise Suite can consolidate an average of 10 business apps used by midmarket companies. QuickBooks Online Advanced ARPC for fiscal 2024 was $3,299.

For fiscal 2025, Intuit is projecting revenue growth of 12% to 13% to $18.16 billion to $18.35 billion. Its Global Business Solutions Group, which includes QuickBooks, Mailchimp and Enterprise Suite as well as workforce management tools, will deliver fiscal 2025 revenue north of $11 billion, up 16% to 17% from a year ago.

This post first appeared in the Constellation Insight newsletter, which features bespoke content weekly and is brought to you by Hitachi Vantara.

Laurent Sellier, SVP Product for Midmarket in Global Business Solutions Group, said: "It's common for customers to allocate tens of thousands of dollars, annually for business management software purchases of tens of thousands and even hundreds of thousands more for external support to get new systems set up. We are encouraged by the customer feedback from early adopters and they're telling us they're getting a lot of value by being on one platform, having one source of truth and easy onboarding due to the familiar navigation."

Intuit estimated that there are 1.9 million mid-market businesses in its priority markets. There are 283,000 QuickBooks Online Advanced customers today, up 28% from a year ago. Those customers now have an upgrade path to Intuit Enterprise Suite. In a demonstration, Intuit used a construction company with 25 employees with $25 million in revenue as an Enterprise Suite customer.

Enterprise Suite includes:

• Financial and accounting tools to prepare financial statements and manage intercompany transactions. Intuit also uses AI to automate planning tasks such as cash flow management, budgeting and profit and loss forecasts with dashboards.
• KPI tracking and insights by project and industry. These KPIs also sync with employee payroll and time data as well as historical financials.
• Mailchimp marketing integration to find, retain and manage customers.
• HR features for onboarding processes, healthcare, retirement and workers comp benefits. Payroll tax calculations, deductions and filings are automated with AI to catch errors.
• Accounts payable and receivable automation and reconciliation.
• Access to experts and services for customer success and customization. Intuit Enterprise Suite will have versions for construction, non-profit, service and project-based businesses.

Impact on competition

Intuit Enterprise Suite may disrupt midmarket ERP by providing a familiar upgrade path for businesses that would move toward Oracle NetSuite, Sage, Microsoft Dynamics and other plays on the bet they'd grow into the functionality. ADP is also expanding in human capital management and could effectively follow the same playbook as Intuit.

For Intuit, Enterprise Suite can also give it a way to leverage its unified platform, data quality and generative AI tools. That platform also gives Intuit the ability to move upstream.

“Innovation for ERP – no matter for large or medium or small enterprises – needs to come from a modern platform, an enterprise application platform (EAP) that supports the generic use case of extension, integration and building proprietary automation, enabled by low-code / no-code,” said Constellation Research analyst Holger Mueller. Equally, it requires to build on common data foundation that supports both Analytics and AI. “Getting these two offerings right is critical and Intuit is making good progress on both ends.”

• Constellation ShortListâ„¢ Enterprise Cloud Finance
• Constellation ShortListâ„¢ Services-Centric Cloud ERP
• Constellation ShortListâ„¢ Product-Centric Cloud ERP

Will Intuit ever threaten SAP, Workday, Oracle and Infor? No. But will it prolong the sales cycle to move up to those ERP packages? You bet.

Recent enterprise application research:

• Oracle Expands Its Lead in Cloud HCM With AI, Workforce Management, and More
• Workday Extend Writes Its Next Chapter: AI
• The JD Edwards EAP Comes From Nextworld
• AI and Automation Breathe New Life Into Midmarket CRM Suites
• UKG Writes Its Next Chapter in the Future of Work
• Infor Invests in AI, Automation, Platforms, and Marketplaces With Infor OS

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

BioNtech's InstaDeep, which was acquired in 2023 for about $682 million, has released a series of foundational generative AI models for proteins and DNA and released them on its DeepChain platform and outlined a supercluster called Kyber.

The news, outlined at BioNtech's AI innovation day, highlights how foundational models are branching out into industry specific use cases. In BioNtech's case, its InstaDeep unit is looking to embed AI throughout the life sciences, R&D and drug discovery value chain.

InstaDeep has even created an AI-driven lab agent built on its proprietary data and Meta's Llama family of models.

BioNtech in recent years has been best known for its COVID-19 vaccine partnership with Pfizer. However, BioNtech historically focused on mRNA cancer treatments. BioNtech is betting that AI can drive its drug pipeline for years to come with its acquisition of InstaDeep, which counted Google as an investor. BioNtech and InstaDeep formed a joint AI lab in 2020 and the partnership quickly accelerated.

Ugur Sahin, CEO BioNTech, explained the company's bet on InstaDeep, which also has its own supercomputing cluster called Kyber. Kyber is coming online in Paris and enables InstaDeep to train its own foundational models without the cost and queue involved with cloud computing.

Sahin said:

"Every cancer treatment for every patient is a new battle. Every cancer cell is different. How can we develop treatments that address tumor cells? Cancer is evolving. Cancer is adaptable. This has now become a high-level computational question."

Sahin added that cancer treatment in the future will start with clinical samples from the patient and an analysis of genetic changes in tumor cells that will generate about 4 terabytes of data for each patient. "We need AI, machine learning and algorithms to come to the right conclusions," he said. "AI gives us the opportunity to do that at a much deeper and faster scale."

Is BioNtech a biotech company or an AI company? Both. Life sciences and AI are likely to become symbiotic.

Ryan Richardson, Chief Strategy Officer at BioNTech, said the company is looking to build an "AI personalized immunotherapy platform." The value drivers for the InstaDeep purchase revolved around cost efficiencies from internalizing model training, building foundational models for vaccines and therapeutics and applying AI to drug discovery.

"The primary use case is to embed AI in drug discovery with the ability to combine our therapeutic platforms on one hand, which are very novel, and the AI capabilities that InstaDeep brings to bear," said Richardson. "There is truly profound disruptive potential in terms of developing or discovering new drugs."

• AI 15O's Matt Lewis on GenAI adoption, psychology and life sciences
• Welldoc Chief AI Officer Anand Iyer on AI, data and healthcare transformation

Karim Beguir, CEO and Co-Founder of InstaDeep, said the goal is to work with BioNtech closely to become "a leader in digital biology." Beguir added that for InstaDeep and BioNtech to lead in digital biology his company also needs to be a leader in AI. "The same technology can apply to multiple use cases," said Beguir. "We are leaders in industrial optimization within biology and outside of biology these add up together. The objective is to continue to be a leading power in the world of AI."

Here's a look at what InstaDeep is working on as part of BioNtech.

A supercomputing cluster named Kyber. Beguir said the Kyber supercluster is built on 224 Nvidia H100 GPUs, 86,000 CPU cores, 1.7 petabytes of persistent storage and 400 Gbps RoCE network. The cluster, built on-premises with Dell, totals about 0.5 ExaFLOPs and is one of the top 20 H100 GPU clusters globally.

"We are now able to take all the work that we have built upon over the last several years and scale it up over the next five, six, seven, 10 years," said Beguir.

InstaDeep uses an in-house rack design that's easy to expand with modular nodes that offer consistent performance, cost, power and cooling. Standard designs will minimize costs over time. InstaDeep also tailored its AI software stack to its workloads with open standards.

Beguir said InstaDeep built the supercluster to avoid vendor lock-in and benefit from predictable costs while scaling models. Kyber enabled InstaDeep to train genAI models with more than 15 billion parameters with hardware efficiency on part with the latest Meta Llama 3.1 foundational model.

Bayesian Flow Networks (BFNs), a new class of generative model that uses Bayesian inference to update beliefs about data. BFNs generate discrete data in a continuous way and are better suited for proteomics and modeling protein folding, function prediction, antibody design and sequence generation.

InstaDeep wants to use BFNs to build foundational models based on heterogeneous scientific data to give scientists more flexibility. A model called AbBFN-X is designed to be a multimodal model for antimodels with 26 different attributes jointly modeled.

DeepChain, a platform designed to use AI to accelerate the R&D pipeline, gains new features. DeepChain is getting generative protein models, ProtBFN and AbBFN, and foundational models for DNA, Nucleotide Transformer and SegmentNT. These models, which can be customized and fine-tuned, are available on Hugging Face under the genomics tag.

• Enterprises leading with AI plan next genAI, agentic AI phases
• 13 artificial intelligence takeaways from Constellation Research’s AI Forum

Laila AI agents built on Meta Llama 3.1. Laila is integrated throughout the DeepChain platform and can recommend models and analyze data with internal and external tools. Laila can also visualize results, plot data and zoom in on certain DNA sequences and positions.

InstaDeep executives said that Laila, which comes in multiple sizes, is more than a chat bot and can use its expert knowledge of biology to reason, make decisions and provide feedback.

The company is also working to leverage its models across scientific and R&D workflows. InstaDeep has designed AI tools to automate labs, annotate tissue, segment pathology images and identify novel therapeutic targets.

Constellation Research analyst Holger Mueller said:

"While most of the CxO attention is on cloud platforms and AI vendors when it comes for the latest on genAI, there is substantial innovation coming from the biotech industry as well. BioNtech (where the founder would go on vacation with its workstation) acquired its own AI startup and it is showing significant progress on what matters at the moment - the 'uber ai' that chooses the right AI / statistical models for positive outcomes in protein folding, cancer research and more. It's good to see more AI model competition, especially coming from a practitioner."

More:

• Themes from the healthcare data, AI, disruption front lines
• 8 takeaways from Constellation Research's Healthcare Transformation Summit
• Medtronic aims to leverage medical device data for AI-enabled care
• Google Cloud's healthcare push accelerates with Vertex AI, MedLM, Healthcare Data Engine
• CVS Health’s Transformation Rides on Data, AI and Customer Experiences
• AI Meets Mental Health Assessments: How Futures Recovery Healthcare, Aiberry Co-innovated

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

Accenture has formed a Nvidia Business Group that will deploy agentic AI using Nvidia's full stack. The move puts some systems integrator heft behind Nvidia's software ecosystem.

As noted during Nvidia's recent second-quarter earnings call, the real competitive moat around the chip giant's business is its software ecosystem. That ecosystem also enables Nvidia to sell its GPUs and AI accelerators. Nvidia made a series of announcements highlighting its ability to leverage its software ecosystem to boost performance. In addition, Nvidia is looking to make it easier for enterprises to bring generative AI projects from pilot to production.

• Nvidia drops NVLM 1.0 LLM family, revving open-source AI
• Nvidia highlights algorithmic research as it moves to FP4
• Nvidia launches NIM Agent Blueprints aims for more turnkey genAI use cases
• Nvidia shows H200 systems generally available, highlights Blackwell MLPerf results

In a statement, Accenture said its Nvidia Business Group will include 30,000 consultants trained on the chipmaker's stack. Accenture's AI Refinery platform will also leverage Nvidia's architecture and foundational models. Accenture said it has booked more than $3 billion in generative AI business. 

Specifically, Accenture will combine its AI Refinery and Nvidia's AI Foundry, AI Enterprise and Omniverse to focus on process optimization, simulations and sovereign AI. The latter has been identified by Nvidia and other infrastructure players like Oracle as a hot market.

The companies will also use Nvidia NIM Agent Blueprint for virtual facility robot fleet simulation and combine it with Eclipse Automation, an Accenture unit that focuses on manufacturing automation.

Nvidia's uncanny knack for staying ahead

Accenture said AI Refinery will be available on all public and private cloud platforms. Accenture said it has used its AI Refinery and AI agents to cut manual steps in campaigns by 25% to 35%, save 6% and increase speed to market by 25% to 55%.

More on agentic AI:

• Enterprises leading with AI plan next genAI, agentic AI phases
• With Salesforce push, AI Agents, agentic AI overload looms
• 13 artificial intelligence takeaways from Constellation Research’s AI Forum
• Boomi CEO Lucas: AI agents will outnumber your human employees soon