After a remarkable financial rebound, Anaplan faced its next challenge - the “ SaaS Apocalypse .” Charlie Gottdiener, chief executive officer of Anaplan, believes the SaaS Apocalypse is overblown. Rather than adversely affecting all enterprise software companies, it will sort the market into winners and losers.

Anaplan is a large player in the enterprise market, one of the relatively few enterprise software companies that generate over a billion dollars annually. Anaplan offers a cross-functional scenario planning and analysis platform whose applications span finance, supply chain, human resources, and sales and marketing. One of the reasons customers like Anaplan is that it provides scenarios that bridge different functional areas in ways other enterprise applications struggle to.

Anaplan was taken private in June of 2022. Gottdiener became CEO in December of that year. The company had grown quickly but was unprofitable, with margins too low to ensure its survival.

Today, the company is dramatically different. At the time it was taken private, the company had $600 million in Annual Recurring Revenue. Despite being a cross-functional platform, the company, Gottdiener said, had a client base that consisted mainly of chief financial officers.

Last year, Anaplan had an ARR of over $1.2 billion and was highly profitable. The company’s improved margins and continued growth have not come at the expense of product development. Significant investments are being made in applications, the platform, and AI.

Gottdiener’s background is business strategy. An obvious step was to lower operational costs, which they addressed by expanding their offshore engineering capabilities.

But Anaplan was trying to do too much, wasting resources in the process. The new team set out to create more value for customers and made that the focus of the business.

They talked to their customers, who liked the platform's flexibility, but said time to value was an issue. Implementations took too long, in part because building the models was complex. Further, ongoing professional services were often required to maintain the model, particularly when it was extended beyond finance to other domains such as supply chain and workforce planning. Customers also told them that transforming data and moving it into the platform was difficult. Anaplan has focused on solving those problems.

The strategy also led them to focus on high-frequency use cases. In the supply chain, for example, those use cases were demand planning, inventory planning, and sales and operational planning. They built out-of-the-box functionality with embedded best practices around those high-frequency use cases.

A small acquisition of a division of a company called Vuealta accelerated this. The acquired division had out-of-the-box demand, supply, inventory, and S&OP that Gottdiener referred to as “pseudo” applications. These applications were built on the Anaplan platform. “We tested those with customers. We said, ‘is this something that you would find valuable?’ The answer was overwhelmingly ‘yes.’”

To make these into full-fledged out-of-the-box applications, best practices needed to be embedded in them; they had to be quickly implementable, they had to be upgradeable, and they needed to be extendable - capable of being connected to other use cases without customization. This framework is now largely automated. Anaplan believes the ability to extend without customizing is rather unique in the market.

“We have a big business around this now,” the CEO explained. “It's really been one of the main growth drivers of the company. We have 28 applications that we've launched across finance, supply chain, workforce planning, and sales performance management, and we're building more.” Anaplan’s applications are largely horizontal; its partners are building vertical applications on their platform.

And just when things could not look rosier, the SaaS apocalypse occurred. Public SaaS enterprise software companies have seen their stock prices plummet, and both private and public companies are facing slower growth. This is because of claims that SaaS suppliers will be disrupted by Generative AI. The ability of Anthropic, OpenAI, and other generative AI solutions to rapidly generate software code will lead companies to build their own applications, thereby avoiding costly SaaS fees.

Anaplan’s CEO does not agree. “Companies that have certain characteristics in the SaaS space are going to do well.” Anaplan’s cross-functional scenario-planning and analysis platform performs deterministic calculations. Generative AI, AI based on large language models, is poor at this. LLM AI provides probability-based answers. LLM AI can be used to improve its solution set. But it will not replace them.

The answers provided by the Anaplan engine must be 100% correct. Large Language Model AI can’t provide this level of accuracy. In the supply chain realm, for example, a company that ships inventory to the wrong distribution center 5% of the time is not doing well. Gottdiener exclaimed, “That would be a terrible outcome!” Similarly, a financial report that looks great but is riddled with errors won't cut it.

In contrast, the CEO claimed, “Our deterministic calculation engine delivers 100% right answers 100% of the time. “An LLM can create great value for customers,” but it's really the combination of LLM and a deterministic engine, plus the ability to provide answers at scale, that creates value.

Further, “as people try to do things that their application software does for them today with large language models, they're going to find that it's highly inefficient and expensive from a token consumption perspective.”

Using Anaplan, companies run multiple scenarios. Every time you run a scenario in a large language model, you're consuming tokens at the request and then more tokens when an answer comes back. “That token consumption is going to be extraordinarily expensive to do things that our platform does at a very low marginal cost, almost zero.”

Gottdiener has explained his thinking in greater depth in a commentary he penned for Fortune . “The interface layer — the polished UI moat most SaaS vendors spent the past two decades perfecting — is being commoditized by large language models. Any vendor whose core value is ‘we make data beautiful and easy to query’ is now competing with LLMs that do the same thing for free, via natural language. Any vendor whose primary value proposition was making data easy to see, interact with, or visualize is in immediate danger.”

AI based on Large Language models “fail at multi-step algebraic manipulation, complex modeling, and any task requiring precise step-by-step reasoning.” But the engines behind Anaplan and other enterprise planning engines also rely on clean master data. “Core systems of record like Human Capital Management and Customer Relationship Management,” Gottdiener wrote, “where elements including hire date, comp number, and closed-won deal amount are governed facts, not suggestions,” and not in danger from LLM solutions.

The LLM does become the user interface. Anaplan built a chatbot in the early days of the Gen AI revolution. “It's got its own limitations,” Gottdiener commented, “but it works. It allows someone to ask natural-language questions and get some insight from our application.”

“Where we're going is building out agentic products.” These will be highly specialized agents that sit on top of Anaplan’s applications, capable of performing tasks autonomously for users. In lower-risk situations, the agents can be fully autonomous. By the end of the year, the company will have over 100 domain-specific, role-specific agents. The first agent, a supply chain agent, will be demoable next month.

Understandable and Auditable Answers

Most users know that LLMs don’t always provide accurate answers. But enterprise vendors also face a challenge: if users are to trust an engine's output, the output must be understandable.

A data ontology layer is important for this. A data schema is needed to transform data, retrieve data from the exact right locations, and deliver answers to exactly where they need to go. For agents to perform with precision, this semantic layer is critical.

Anaplan will invest in knowledge graph technology. This graph will map the end-to-end data flows and define their rules. This will be particularly important for analyses that cross enterprise domains – finance, supply chain, human resource planning, and so forth.

For example, if the demand forecast (supply chain app) shows that the quarterly revenue goal won’t be met (finance), an agent could suggest a trade promotion (marketing). Another agent could assess whether sufficient factory capacity and warehouse space (supply chain) are available to make that promotion actionable. The data graph helps the cross-functional inputs and outputs move correctly.

Agents and LLM AI Must Work Together

Arguably, the agentic revolution is more important than Generative AI for enterprise applications. The agentic enterprise can be used to answer the "why" questions: why did something change? Where did they change? What caused the change? And then, based on the answers, to dynamically run the right kinds of scenarios.

Agents can also be used to monitor and act. For example, a planner can tell an agent, “I want you to constantly monitor what my lead time metric from this supplier is.” This is a core input into the inventory safety stock calculation. If the lead time has changed, the agent can be instructed to correct it, thus ensuring more accurate inventory levels.

But when it comes to explainability, generative AI and an agentic framework can work together. When planners get an optimization or forecasting answer they don’t understand, they have to manually drill into the application to understand how the engine generated it. They often fail.

Anaplan’s agents are being designed to provide the decision framework. The agent can tell the planner where the source data came from, which parameters the calculation was based on, the decision-tree logic, and, for the mathematically minded, even the formula.

LLMs are very good at explaining what's changed. So, when a planner runs different scenarios and is looking at these scenarios next to each other, LLMs can do a great job of saying, "When you changed the demand assumption, the answer that came back was based on giving you more demand for these products. This is what caused a capacity issue at this factory.” It will be difficult, however, to get to full explainability for supply planning optimization. This is the classic black box application .

Right now, this is interactive and conversational. “We think that's not the right way to do this,” the CEO commented. The human brain needs the right context to ask the right question. The context might be an event alert, a trending analysis, or a text box that says, “pay attention to this.”

“This is the most dynamic time of my career. It's terrifying, and it's invigorating at the same time.” Nevertheless, Gottdiener concluded, “there are a lot of reasons for us to be optimistic.”