After the Max: The Evolving AI Pitch

A few weeks ago, Salesforce made an unprecedented announcement to investors. Their profit margins would remain unchanged for the next quarter, because they spent too much on Claude. To be clear, this message is largely spin to promote the new Claudeforce partnership. However, it does fit into a narrative that has been building over the summer: AI budgets are too high.

Cost vs Power

Tokenmaxxing was never going to be sustainable. It had a time and place. Unlimited tokens allowed engineers to experiment with a highly controversial technology, which has transformed software development pretty much overnight. By encouraging broad usage during the initial gold rush, tech firms ensured their workforces became comfortable with AI assisted coding. Now they're more interested in control costs. AI labs have noticed. 

Ever since the leading AI labs announced their IPOs, the overarching narrative has shifted to one of token pricing and usage limits. For the last few releases, much of OpenAI’s messaging has been around cost instead of capabilities. The over the top reaction to the recent ChatGPT Astra launch was symbolic of this. Astra the first OpenAI model to be promoted for its AGI capabilities this year. The previous wave was pitched as a cheaper alternative to Anthropic, whereas the new model is pitched as more powerful than Claude Fable.

No More Vendor Lock In

The changing marketing pitch has exposed the weakening moat within which the leading AI labs are operating. A Ramp report released last week backed up the twin trajectories facing Anthropic and OpenAI. In general, usage is increasing. Business users are experimenting with longer queries and scheduled tasks. These are the starting blocks for the agentic future that labs have been building towards. Yet, light AI users are not loyal to a particular model. They often pick the most convenient AI vendor, and are much more willing to rework prompts for different models. It's a market where convenience matters as much as output.

There is an exception though. Trouble is, these users aren't any more loyal to a particular lab, but for a different reason. It's also the only part of the AI market which is seeing declining revenues. The 1% biggest AI users are cutting back their LLM usage. These are the frontier firms. Lured in by the promise of vibe coding, software companies have adopted AI at far higher rates than anyone else. Yet, this hasn't translated into more reliable code or quicker release cycles. With coding assistants now firmly embedded in software development workflows, dev teams are cutting back token consumption in order to control costs.

With ROI under scrutiny, token efficiency is just as important as token utilisation. That's true even for projects where AI has demonstrated ROI, such as vibe coding. Many of the biggest AI users are experimenting with cheaper open source models in order to reduce token consumption further. Some tech firms are trying to train their own models; others have adopted Chinese models such as Kimi or Deepseek as part of their search for efficiency. Politically motivated warnings about the risks of overseas models should be seen in that context.

Beyond Software

Fortunately for AI labs, developers are only part of the overall market for large language models. Agentic AI is still in its infancy. When it comes to AI adoption, business teams are still catching up with their engineering colleagues. In most sectors, user friendly automation tools are still relatively new. Claude Cowork and ChatGPT Codex have only been available for a matter of months. Learning to work with them takes much longer because ROI generating use cases for Cowork are much harder to define. 

Marketers have an advantage in that respect. Content creation is such an obvious use for generative AI, and many businesses have already invested in tools or workflows to support AI-assisted copywriting. However, CMOs are struggling to measure the ROI from in-sourcing content production. From a finance perspective, it's just shifting content budgets elsewhere. Better sales enablement through AI SDRs has become a major focus area for this reason, because it allows fewer sales reps to close more deals.

There's a lot more that marketing ops can do to embed AI within marketing, even if it's just embedding the technology into existing campaign setup or data management workflows. Doing this extends marketing automation into the types of one-off situations that needed extensive manual workarounds in the past. The challenge has been linking the AI into those existing process flows, and then giving sufficient context to make AI useful. Improved tooling and enhanced MCP integrations are driving experiments with AI automation. Actually turning those experiments into production-ready agents is a much more complex task. 

Marketo Launches New UI

Both Marketo and Eloqua are getting a fresh lick of paint. Adobe and Oracle are currently running UX beta programs. The changes are purely cosmetic: a simplified design language with new icons and new colours. At the moment, the updated user interface is being rolled out on an opt-in basis. It's worth trying out though, both for familiarity and to provide feedback to the developers.

Operationalising ABM in the Real World: Your Step-by-Step Guide

Finding it difficult to navigate the complexities of ABM. A new eGuide from CRMT Digital cuts through ABM hype, explaining how to implement and run an ABM program in practical terms. Boost ROI with practical approaches to account-based targeting, personalisation and engagement.

Marketing Operations Roadmap Matrix

A Marketing Operations Roadmap is an essential strategic tool that aligns marketing priorities with your overall business goals. Any roadmap begins with understanding where your marketing organisation is right now, and then defining where you want to get to. This free Marketing Operations Roadmap Matrix from CRMT Digital allows you to do exactly that.