The average enterprise has more than 100 SaaS licenses. A new wave of LLM-enabled functional software will almost certainly swap out many of those providers.

But we think LLMs will also drive a new set of customizable platforms that allow customers to build their own workflows designed specifically for their organization. This category is called composable software.

Over the coming years, we believe that a substantial portion of enterprise software spend will shift from out-of-the box products to customized workflows built on composable software. We are excited by the opportunity to build these platforms in different domains.

What is composable software?

Composable software provides core infrastructure and a platform for customers to design, implement, and modify their own workflow software.

Examples of earlier composable software platforms include:

• Retool: Offers lightweight interfaces and components to manage and manipulate your data for internal or external applications
• Zapier: Allows you to build interfaces, perform actions, and sync data across applications

Why now? Composable platforms will unlock LLM-enabled automation

As discussed in our previous post, LLMs enable a new level of workflow automation that the previous generation of tools could not. These capabilities will dramatically change how many functions operate.

But LLMs make analyses and perform actions based on data. Applications will need access to data across the entirety of the workflows they automate, ideally with consistent content and structure. It’s clear why an LLM automation application will be most effective when built on a single platform that provides consistent APIs and data layer, versus one that must integrate with a dozen applications owned by different vendors.

We believe the long-term capabilities and value of LLM-based automation will be what drives a consolidation of software to these composable software platforms.

There are a few other reasons why composable software platforms are now more attractive:

1. Modern application services: There are modern API-based services that provide functionality that was previously hard to build (e.g. text messages, payments, verification)
2. Modern data infrastructure: Storage and compute is cheap and performant. Most companies can easily dump all of their data in a modern warehouse, and store, duplicate, and process it without worrying about needing to hyper-optimize code.
3. LLMs as implementation tools: LLMs will drive many of the application workflows, but also are an extremely useful tool for implementing and updating composable software platforms themselves. Customers can use LLMs to help write the code that expands or modifies their software’s functionality.

What makes a composable software platform successful?

Domain-specific

The companies building composable software today like Retool and Zapier tend to be general. We believe that there are exciting opportunities for new composable software companies that focus on particular domains. These domains could be:

• Verticals: e.g., manufacturing, healthcare, financial services
• Business functions: e.g., supply chain/logistics, security finance/accounting, RevOps

Focus will allow new companies to better compete with the incumbent specialized software providers in their domain. They can design a platform and data structures that are relevant for business logic and workflows. They can also build domain-specific workflows and templates that customers can use to immediately realize value.

Data model and querying layer

The core component of any composable software is the data model and a development framework/language to make use of it.

Our hypothesis is that the most successful composable software platforms will create an opinionated ontology and data model. The way you’d design a data model for a broad composable software platform is very different from what you’d need in a vertical application.

The data model is the hardest part to build right. It should be simple enough that customers can easily spin up a new application, but detailed enough to handle the intricacies of a specific domain. It should be usable out of the box, but fully extensible if a customer needs to modify core concepts or schemas to meet their business needs.

Time to value

A common challenge for composable software platforms will be time to value. When you buy a SaaS product, it’s often ready to use immediately. If composable software requires weeks of work to implement, many customers won’t be interested.

There are a few ways successful composable software platforms reduce time to value:

• Narrow initial use case: Target the initial use case to a specific team and workflow to demonstrate value while limiting data and system integration requirements.
• Templates: Provide pre-built configurations that can be used as-is or as starting points for modifications.
• Product/GUI: Offer low/no-code tooling that enables non-technical users to create and modify workflows.
• Automation: In the longer term, composable platforms can use ML or rules-based systems to evaluate a customer’s needs and automatically generate configurations.

Impact of composable software platforms

For buyers

Composable software can certainly save costs. Specialized software providers charge a premium because there are limited alternatives.

Building on a composable software platform, companies can recreate existing functionality at a fraction of the cost. They can also simplify procurement and vendor management.

But we think composable software will be much better than existing vertical software.

Companies can tailor software to how their business operates. They can design more powerful or efficient workflows that match their team’s processes. It’s easier to onboard new applications and users with a consistent design language. And companies can extend and modify any application themselves, adding a new feature or workflow in days instead of months of professional services.

Composable software can connect systems that were previously incompatible or prohibitively expensive to integrate. Teams can build workflows to generate financial reports based on ERP data. They can use customer demand data to update operations in SCM software.

As discussed above, software is also the right foundation for automation. With unified workflows, data structures, and domain-specific language/interfaces, teams can build LLM systems for autonomous and human-in-the-loop automation.

For startups

Composable software is not the easiest product to build. It takes large upfront investment in infrastructure. It also requires customers take a leap to switch from purpose-built software to a more generalized platform they must configure themselves.

But if successful, there are opportunities to build massive companies in these spaces. In addition to consolidating spend that was previously spread across multiple software platforms, composable software (+ LLM automation) can create massive new efficiencies for businesses. And like SAP or Salesforce, these systems will be extremely sticky once broadly implemented.

These dynamics are very well suited for a land and expand approach. Companies can start by selling a specific use case, expand to adjacent applications, and eventually build a suite strategy that allows them to capture value from multiple functions.

Examples of composable software platforms

Accounting

While all companies have a general ledger like NetSuite, the enterprise accounting space is dominated by vertical software. Blackline and FloQast do monthly close/reconciliation. Workiva does reporting. Avalara and Vertex do tax compliance. Building workflows across these involves manual copying, pasting, and manipulating data in Excel.

We think new companies can displace each of those vertical software companies with a modern data infrastructure and LLM-powered automation.

But thinking more broadly, all of these tasks are made of similar building blocks: extracting, transforming, comparing, and summarizing tabular and textual data. There may be an opportunity to build a general accounting platform that allows teams to build their own workflows and automations. This could help accounting firms serve more clients, power a tech-enabled services firm, or streamline an internal accounting team’s operations.

Supply chain

In supply chains, the movement of data is as important as the movement of goods. This has always been a pain point in industry.

Large enterprises set up integrated ERP systems like SAP, which are expensive, complex, and slow to implement. Business data sources (e.g. customer, marketing, IoT) and applications (e.g. BI, customer apps, internal tooling) have proliferated, but making any changes to the core ERP infrastructure or workflows requires months of custom work.

The Electronic Data Interchange (EDI) interface used to share data across companies is non-standardized and brittle. The result is that a majority of companies find themselves sharing key information in emails, PDFs, and comment fields.

With composable software, supply chain teams can define workflows and processes that are bespoke for their business. They can connect parts of the business that typically can’t share information (without expensive integration efforts) – e.g. linking inventory management with CRM/demand forecasting. As the business grows, teams can update data structures, workflows, and automation steps on their own to be flexible to evolving needs.

Data pipelines

Today, data integrations are built with a web of custom code. Even using integration platforms like Fivetran or Airbyte, pipelines require transformation and querying logic. With unstructured and semi-structured data (emails, PDFs, spreadsheets), the problem is intractable for most companies to tackle. There have been a number of low-code data pipeline platforms, but they haven’t worked well in these complex/messy data domains.

One of the most powerful capabilities of LLM systems is that they provide robust extraction and transformation functionality. We think this will for the first time enable true composable data pipeline infrastructure, allowing technical and non/semi-technical users to build, maintain, and modify production data ingestion pipelines that can power their business and workflows.

—––

If you’re building a composable software platform (or something like it), we’d love to hear from you at info@theory.ventures!

The most compelling value proposition of B2B LLM applications is workflow automation. In nearly every function, headcount costs are an order of magnitude larger than software spend. Outside of cost savings, AI automation can improve the speed and quality of the work performed. An AI-powered customer service bot means a customer will never have to wait for an agent. An AI-powered security analyst can work around the clock, and won’t get bored after looking at 100 nearly identical alerts.

But Robotic Process Automation (RPA) tools have been around for decades – both general platforms like UiPath and Zapier, as well as function-specific ones like Workday (human resources) and Demisto (security). So what’s changed?

Historically, RPA just hasn’t worked well enough to automate end-to-end workflow. Their “if this, then that” type functionality can tackle some small tasks. But they’re too cumbersome and brittle to handle the vast majority of work that people do on a daily basis.

LLMs change this equation. They enable systems that can actually accomplish human work, instead of just moderately increasing efficiency. This will result in a new age of automation software.

Business workflows require human-like decisions and judgment

While many business workflows follow a consistent high-level process, you can’t blindly follow a sequence of steps robotically. Humans running the process make many decisions and judgments along the way.

For example:

• Someone reading an email that mentions “invoice” must determine if the email is sending an invoice, requesting it, or referencing a previous one.
• An employee entering a purchase order will quickly notice if the quantity is off by a factor of 10 compared to previous orders, and flag it for review.

It’s clear why traditional automation platforms fall short. Enumerating all of the different possibilities would require hundreds or thousands of rules and workflows, with exception handling for every possible edge case. Any change to the business process likewise requires rewriting all of these rules. Many of the judgments require semantic understanding and human-like logic, which isn’t possible with rules-based systems.

Integrations and interfaces are complicated to build

Most valuable business workflows require read and write access to different systems and software. Many involve both structured and unstructured data, with text inputs via email, forms, or chat interfaces.

RPA platforms provide integrations to many other pieces of software. But for most, these interfaces aren’t easy to use. They require a developer to implement. They are limited in functionality by the API of the third-party platforms. Any access to data requires customized queries and transformations. Unstructured data is almost impossible to work with.

When any of these things changes, the whole process breaks.

For example:

• When security operations centers (SOCs) investigate alerts, they need to pull data from a variety of systems – SIEM, NDR, EDR, threat intelligence feeds, etc. Each one of these has its own data structure, querying language, and field parsing.
• Supply chain communications use the Electronic Data Interchange (EDI) format, but it is a weak standard whose structure varies by company and even employee – for example, someone might put information in a “product” field, others in a comment.

It doesn’t matter how simple it is to design a workflow in a GUI: If the pipelines and integrations to power it take months of specialized work to build and maintain, it makes many projects infeasible.

Modern LLM systems solve these problems

LLM systems solve both of the challenges above.

LLMs can make many small judgments and decisions effectively, using semantic understanding plus basic logic/reasoning. They can categorize a message or decide if a question has been answered.

LLMs can also be used for ETL and interface, e.g. extracting information from an unstructured email or writing queries to access different data sources. This is still an emerging capability, but is rapidly improving.

We do not need fully-autonomous AI agents to realize the benefits of LLM-based automation. We expect the majority of applications will still be built with a concrete process description and guardrails designed with expert knowledge.

What the LLM unlocks is the human-like reasoning that allows a system to navigate the branching points, reason, and provide flexibility that is not possible with rules-based systems alone.

Examples where LLMs will transform workflow automation

Security operation centers (SOCs)

SOCs staff teams of analysts and engineers to triage, investigate, and resolve alerts.

Security Orchestration, Automation and Response (SOAR) platforms promised to automate security workflows. But the rules-based playbooks are only good at simple, repetitive tasks like “send a Slack notification when you receive a high-severity alert from X system” or “email Y users daily until they change their password”. They cannot perform the actual investigation, data gathering, and analysis that humans do to determine if incoming alerts are real or false positives.

With LLM-based automation, platforms can replicate analyst workflows like a human, querying data from different systems, interpreting it, and executing actions to mitigate the issue. These systems can respond much more quickly to alerts, work 24/7 without alert fatigue, and allow humans to focus on higher complexity issues.

Supply chains

Supply chains are built on complex, legacy infrastructure. Large enterprise ERP systems like SAP and NetSuite can take months and millions of dollars just to implement.

Because of this, a huge portion of supply chain data is exchanged via EDI and unstructured documents like emails, PDFs, and spreadsheets. These have limited the opportunity to build automation.

LLMs make data extraction and transformation much more robust. With these capabilities, companies can automate end-to-end workflows – not just to save labor costs, but to drive better supply chain operations and provide better service for customers.

—––

If you’re building a company around LLM-based workflow automation, we’d love to hear from you at info@theory.ventures!