AI for Business Users is Just Starting — Focus on High ROI Use Cases, Architecture, and Gross Margin

AI for Business is still in its early days. I’ve talked with a lot of AI for Business entrepreneurs and cloud executives who are asking themselves if they should start a pure play or native AI company, add an AI extension to their product line or to join a Business AI startup. My advice: focus on the highest ROI use cases, evaluate the AI architecture, and dig into the gross margin before starting a company, setting customer expectations or making a career move.

So far, most of the AI progress has been focused on the infrastructure side. Consumer applications of AI have far outpaced business use cases to date. These companies and applications are being fueled by lots of venture capital and corporate research spending due to the high computing cost of AI and the increasing complexity of the workloads.

Now, a shift is beginning. Business users and teams are starting to actively engage with AI at the user and teams levels. This adoption has been proceeding slowly with pilot deployments, largely due to concerns about ROI and expense, legal and IP issues, the potential for AI hallucinations, and AI steering committees.

But it is steadily progressing nonetheless. A few weeks ago, at the Salesforce global user conference in San Francisco, AI for Business took a significant step forward with the unveiling of Agentforce, Salesforce’s new suite of autonomous AI agents. Agentforce demonstrates the great potential of AI to improve business outcomes. The platform enables users to build, test and supervise AI agents that are fully integrated into their Salesforce instance. These autonomous agents can be customized to support customers around the clock, handling tasks across the service, sales, marketing and commerce departments. Salesforce’s commitment to AI sends a powerful message to business users and other enterprise software companies about the future of the technology in the business world.

Of course, some industry luminaries, like Andreessen Horowitz, think Salesforce isn’t going far enough and that software companies built from the ground up with AI will ultimately displace the incumbents. In a recent piece, they argue that “AI will so fundamentally reimagine the core system of record and the sales workflows that no incumbent is safe.”

The cost — and value — of AI for Business

Time will tell if Andreesen Horowitz is right. But what is clear is that the starting gun has been fired. AI for Business is off and running. Will companies fully embrace it? Many will. But many will wait, because AI infrastructure and applications can be very expensive to buy for the customer — and very expensive to run for vendors.

How expensive? Let’s look at the pricing. Salesforce, which typically charges $100-$200 per user per month, introduced its first AI edition earlier this year at $500 a month — a substantial increase. The reality is that the development of AI tools is cost-intensive and vendors have to charge high rates to recoup their investment. To make the price more palatable, Salesforce announced a new per-agent conversation pricing model, charging $2 per interaction. This is not unreasonable, given that a phone call to resolve an issue might cost $8. Nevertheless, those rates are going to cause some sticker shock in the customer base. So vendors must identify high-value use cases where the ROI for the customer justifies the elevated price point.

For executives at cloud business application companies who want to add AI to their product suite — or outright join an AI company — it’s crucial to assess whether the value that an AI application provides is worth its high cost and to really understand if the underlying AI architecture delivers accurate results and is profitable for that use case.

It should be noted that, unlike traditional cloud applications, which often have 70%-80% gross margins, AI applications running on name-brand Large Language Models (LLMs) have lower margins due to their high computing costs. All well and good. But if the value delivered by AI applications is incremental rather than transformative, many customers will balk at paying a substantial premium, and the AI for Business company will have a hard time making a profit.

Justifying a big investment in AI for Business applications

To convince customers that their AI applications are worth the price, enterprise software companies should focus on identifying the highest-value use cases. Instead of offering AI packages outfitted with all possible capabilities — and a price tag to match — enterprise software companies should assess their products from the customer use case and ROI perspective, and then align price points with the value they give the customer.

It’s important to say here that most current approaches to AI for Business are not economically viable. Many customers are not generating sufficient savings or returns to justify a big investment. And if customers aren’t seeing enough value, vendors won’t either due to the high computing costs. So, it’s a bad outcome for all parties involved.

Yes, AI LLM infrastructure costs are expected to decrease over time. But they’re not dropping as quickly as many had hoped, in part due to the increased complexity of the workloads and the speed that users want results. This means executives looking to add AI to their company’s offerings must be laser focused on high-value, mission-critical use cases that are true differentiators rather than mere nice-to-haves.

Meanwhile, those executives considering a move to a pure-play AI for Business company should examine the company’s offering. What’s the cost to solve a specific problem? What’s the potential gross margin of the offering? Ask whether the problem the company is addressing is valuable enough to justify the cost of developing AI applications to solve it. Some companies, despite having incredible teams and initial success, will inevitably struggle with poor gross margins that are difficult to overcome.

Small Language Models — SLMs: one answer to the high cost of AI

The Business AI industry is now seeing a shift toward small language models (SLMs) from LLMs. This is because SLMs offer several key advantages over LLMs. One of the most significant is that SLMs tend to produce fewer hallucinations and higher accuracy rates because they’re more narrowly focused. SLMs are also far more computer resource-efficient. They enable companies to deliver high accuracy and ROI without the hefty infrastructure costs that come with developing and maintaining LLMs, which is good for the customer and provides a good gross margin for pure-play business AI companies. Furthermore, SLMs can be deployed on-premises or in private clouds, allowing businesses to keep their sensitive data secure within their own environment. This is particularly important for industries with strict privacy regulations or concerns about intellectual property.

4CRISK.ai: The AI for Compliance Company

A good example of a pure-play, native AI startup effectively using SLMs is 4CRisk.ai. Rather than creating a general-purpose AI company, their mission is to revolutionize GRC from Risk and Compliance Officers with specialized SLMs. Their native GenAI platform tackles previously insurmountable challenges: real-time compliance monitoring and proactive risk identification in the face of a constantly evolving regulatory, business, and external risk environment.

4CRisk’s founders launched their company utilizing AI, ML and small language models before ChatGPT was unveiled. Their application suite brings in all the applicable laws and risks in the world and marries them with a company’s internal control environment to answer these critical questions that were never possible. They then added conversational AI capabilities that allow business and compliance people to ask compliance-related questions and get answers in seconds rather than days or weeks.

4CRisk is revolutionizing a manual process that large companies might pay $30 million a year to outsource and is reducing those compliance costs by around 50%, while also decreasing the chances of incurring compliance penalties, fines and legal expenses. Its focus on a specific, high-ROI problem, combined with the founders’ deep compliance domain expertise, has enabled 4CRisk to develop a groundbreaking solution using an SLM architecture that delivers 70%+ gross margins, which will continue to increase as they add new products.

Their early pilot enterprise customers have moved to large scale production deployments with high accuracy, exceptional ROI, and meaningful annual recurring revenue (ARR) contracts with lots of future upsale opportunities.

How the AI for Business market leaders will be chosen

AI for Business users are now moving from pilot deployments to full-scale production implementations. The value of these real-world deployments and references cannot be overstated. They serve as powerful proof points that AI can deliver real business results — and deliver them today, not tomorrow.

The key point is that production references demonstrate ROI and what’s possible. For decision-makers evaluating these solutions, the ROI is significant, but so are the risks if things go wrong — making vendors with proven production references and successful customers the clear choice.

The AI revolution in business is not just about providing flashy new technology, it’s about showing that the technology can drive meaningful outcomes for customers, improve efficiency and create competitive advantage. Business AI companies that recognize this — and develop high-value, results-driven AI solutions — are the ones that will emerge as the true leaders in an increasingly crowded market.

~ Matt Holleran, General Partner, Cloud Apps Capital Partners