Unsurprisingly, AI has been a key discussion point for public SaaS companies in their earnings calls over the past year. AI was mentioned 319 times in the last 2 quarters across 4 companies I arbitrarily picked—Salesforce, Freshworks, ServiceNow, and Gitlab. 

We’re still in the early days of AI, with a lot of unknowns and experimentation. What seems to be changing in the last quarter, is that AI, specifically generative AI, is going from a buzzword to capabilities being released to customers. Executives are starting to factor in tailwinds from AI into their forecasts. 

These incumbents have multiple advantages of scale—distribution, data, and a broader product footprint in an economic environment where companies are looking to consolidate spend. 

The typical answer to how startups win is speed. Iterate faster, move faster. 

Now, it seems that incumbents are innovating at breakneck speed. They’ve gone from nothing to launching new generative capabilities to customers in a few quarters. I’d argue that the innovation is typically incremental and safe. Their pricing is mostly unchanged. 

Let’s dive into some observations on how public SaaS companies are rolling out AI to their customers:

The AI Use Cases are Fairly Incremental and Innovation so far is Limited

Today, the generative AI use cases for application software companies center around summarization: pre-fill responses to a ticket, summarize a document or answer a question from a knowledge base. 

Most of this can be achieved via RAG or prompt-tuning to a foundation model. Companies may have the logic to use different models across use cases to balance accuracy, performance, and cost. Nonetheless, it’s unlikely to be a defensible source of IP. This is evident from how quickly companies were able to build and roll out these capabilities to end users. Here is a quote from the latest Freshworks earnings call that underscores this point 

We built generative AI capabilities across our products in just a few months. We started with Freddy Self-service through bots and later introduced Freddy Copilot, built for Freshchat first and then extended to Freshdesk and Freshservice.

Additionally, most use cases are ‘copilots’ with a human-in-the-loop before the end result reaches the end user, offering more tolerance to inherent errors and hallucinations. 

There are multiple caveats here:

  1. This applies to application software companies. Innovations from gen AI are more foundational for infra companies including Elastic which launched hybrid vector search and Snowflake’s launch of Cortex, a platform AI layer and a significant bet for the company.

  2. Let’s not confuse defensible innovation for impact. Even a 5% increase in productivity can be significant in terms of CSAT and competitive advantage. Here is an example from Klaviyo that suggests a 50% lift:

    “(Customer) Willow Tree used some of our AI features to figure out new groups of customers to target with campaigns, figure out pools of customers and the content that was relevant to them. As a result of taking those suggestions that came out of our engine, they were able to increase the revenue they drove through Klaviyo by over 50%.”

  3. Companies are innovating, especially when it comes to moving from ‘copilot’ to ‘autopilot’. Freshworks Freddy self-service bots are one example. The vision for Klaviyo is autonomous marketing campaigns where what, when, and to whom is decided by AI with limited human intervention. This impact can be manyfolds greater, but it’s likely to take far longer.

AI Pricing is Aligned With the Company’s Existing Pricing Structure

AI capabilities are being priced as one of the following:

  1. A new SKU.

  2. Available only in higher tiers.

  3. Usage-based pricing.

  4. Available to all paid customers at no extra charge.

The pricing and packaging strategy for AI is highly aligned with the existing price structure and growth levers. Here are some examples:

  • ServiceNow is an enterprise offering with tailored (opaque) pricing, typically with multiple line items. So its AI offering is a new SKU.

  • Box offers 5 different tiers of its core service. AI capabilities are only available in the more expensive tiers, however with usage caps. Box also offers a ‘platform service’ on a consumption-based model and charges for AI usage above the cap.

  • Elastic has a consumption-based model where they universally charge based on computing and storage across all their product lines—search, observability, and security. They’re adding vector search and other AI capabilities with no change to the pricing model, but expect these capabilities will drive usage and revenue.

  • Zoom’s primary growth lever is moving people from free to paid (and reducing churn), so its AI capabilities are available to all paid users.

  • Freshworks offers a combination of tiered pricing and add-on SKUs. As of today, its AI capabilities are offered as add-on SKUs.

The uplift can be meaningful. A small improvement in free-to-paid conversation significantly improves revenue. The list price of Freshworks AI add-on is priced at £23 or 41% of its pro tier.

As AI moves from efficiency (copilot) to autonomously handling use cases, I expect pricing to move towards consumption-based pricing. For example, Freshdesk Freddy Self-service bots are priced based on interaction versus their core business model of seats. In areas such as support or sales, AI may start replacing users, creating an innovator’s dilemma for incumbents.

Executives are Starting to Factor in Tailwinds From Access to AI, but not Ongoing Usage

As AI capabilities move from Beta to General availability, executives are starting to factor tailwinds into their forecast. Probably the most bullish sentiment was from Bill McDermott, CEO of ServiceNow:

In Q4, our gen AI products drove the largest net new ACV contribution for our first full quarter of any of our new product family releases ever, including our original Pro SKU… There's a real appetite to invest in gen AI, and there's no price sensitivity around it because the business cases are so unbelievable. I mean if you're improving productivity, 40%, 50%, it just sells itself.

Here is an excerpt from BOX Q4 earnings

Since rolling out Box AI in Beta to Enterprise Plus customers in November, we have seen a number of existing customers upgrade to Enterprise Plus to gain access to Box AI.

In their Q4 earnings, Elastic calls out the usage of generative AI not being meaningful since their customers are mostly using experimentally for non-production use cases:

In terms of Generative AI, it has had—it's definitely contributing in terms of revenue, but given our overall size, I'd say the contribution is still in the early days.…. A lot of the use cases are still internal facing as customers are continuing to become more and more confident about their ability to manage the risk of hallucinations.

The example that bucks this trend is Zoom’s AI companion:

Zoom AI Companion has grown tremendously in just 5 months with over 510,000 accounts enabled and 7.2 million meeting summaries created as of the close of FY ’24.

Generally, there is more caution around baking any sustained usage into the guide. Here is an excerpt from Freshworks:

But for now, we're very much focused on adoption. That's a good leading indicator for us. In terms of how we're thinking about it for the full year, we haven't baked in a meaningful upside to the year from AI because we wanted to see how things play out before putting anything into our number.


Startups will still win on speed—especially with their ability to make bold bets. Not just a copilot offering efficiency, but autonomous systems offering a step change in driving outcomes. Often, via a radically different UX or pricing model. They’ll look to sell services or outcomes directly versus just software. Some examples include

  • Cognition Labs that powers an AI software engineer, not a copilot.

  • Pilot offers an AI-powered book-keeping service.

  • Contlo offers an AI-native autonomous customer engagement tool.

  • 11x.ai offers automated SDRs, not software for SDRs.

  • Zowie provides support software with 100% transaction-based pricing, where most incumbents are primarily seat-based.

This is where incumbents with their larger customer base will be slow to change due to natural risk aversion and/or innovator’s dilemma. Also, it’s just harder. There is more technical risk, PMF may take longer especially in trusting AI, and GTM playbooks may need to be more nuanced. At Next47, we understand this, and if you’re a startup pushing the envelope, we’d love to hear from you.