What does the post-DeepSeek era of AI mean for marketers?

I’m still replying to texts from colleagues (and friends, and family members) about DeepSeek’s breakthrough last week, and what it means for the future of artificial intelligence in the Valley. Our data scientists and ML engineers are busy testing DeepSeek, as we do all major models; so this week, I thought we should discuss a more specific piece of that question:

‍

What does the post-DeepSeek era of AI mean for marketers?

‍

First and foremost, we now know the cost of compute power using LLMs is no longer a key budgeting issue. This was enforced days after the news broke, when researchers at UC Berkeley were able to create a small scale reproduction of DeepSeek — for just $30 in compute costs! Of course the final training run for a small scale model build like this (or even DeepSeek's reported $6M cost) is just a fraction of the total infrastructure and R&D costs to catch up with the current market leaders; however, the main point here is the models themselves are rapidly being commoditized.

‍

I'm seeing first-hand the rapidly growing cost effectiveness of LLMs at my own company: In just the last year, our model training costs have been cut in half.

‍

This macro trend probably represents the imminent end of Google Search as we’ve known it over the last 25 years. Now that we can very affordably apply LLMs for most search requests, human use of Google-style landing pages rapidly dissipates. I think it’s likely up to 80 percent of classic Google search queries will be replaced by new AI-agent based search experiences over the next few years.

‍

And these cost savings are particularly exciting for digital marketing with startups and smaller companies. Without having to pay exorbitant computing costs, they can better compete with larger companies, creating LLM–powered content, analysis, and marketing research plans with modest budgets.

‍

The lessening of cost considerations also suggests that marketers should no longer treat AI as a “nice to have if we can afford it” side tool, but as an essential part of their work. Here’s why:

Content Marketing in the Era of Agentic Workflows

With LLM computing costs decreasing, the emerging opportunity for marketers goes far beyond content creation and summarization. My advice is they look into the potential of agentic AI workflows. Instead of using a single LLM for generating outcomes from direct prompts, agentic workflows ask a reasoning AI how it would solve a particular problem – and also, to ask other more specialized AI to solve it.

‍

Agentic workflows can help marketers better and more comprehensively collect customer insights into their audience and their competitors, along with the strength of their content strategies and business intelligence. And in a refreshing break from the SEO era search terms / web analytics mindset, we’re able to explore the market using more natural language definitions and topics analysis, and find breakout trends. And post-DeepSeek, do so at increasingly reduced cost, and with specialized vertical market applications!

‍

At Rellify, we've developed a platform that automatically generates company-specific machine learning-based models with deep knowledge of relevant topics which in turn are used to prompt LLMs. This way, our customers can focus more on the strategic questions like what topics to cover and differentiate their offering ---- focusing the power of LLMs in the creative process to help generate relevant content.

‍

This approach has repeatedly led to our customers' content investments driving visibility in both AI and classic SEO search results, because our vertical models identify what topics matter -- what are the relevant natural language questions that quality content should address -- not just what keywords drive the most search volume.

‍

All of the above is not to imply Google Search and the company's search algorithms are going away, I should add; data analytics around Google’s web index is still important for marketers working with LLMs; my own company uses it as a the best approximation of audience interest. It’s just that we should expect it to become less important – especially when building our marketing / content strategies.

‍

As for our company, Rellify, we have been planning for the time when OpenAI is no longer the undisputed industry leader in LLM models. I believe I’m not the only one glad to say that time is now.

‍

If you’re interested in what we’re doing, please get in touch; happy to chat further!

‍

Peter Kraus is CEO and founder of Rellify, the Content Intelligence company built from the ground up with AI to generate insights and strategies that transform the entire customer journey.