ChatGPT is amazing for one-shot prompting. You can give it a single, brief prompt, and it will complete the task in one, single flow. Humans can’t do that - but that’s actually a good thing, for a couple of reasons. The human way of doing things is less linear: a marketer will brainstorm, pick keywords, then draft. Then, she’ll pour herself a coffee, come back, and redraft.
This takes a lot longer, but the iterative process produces coherence and relevance, which LLMs struggle to match (this article is AI-free, precisely because the nuances of the subject matter were too fine-grained to prompt into an AI tool). This is a familiar problem, and it was only a matter of time before solutions were developed by AI innovators like Intel® Liftoff member PixelML. In a recent Intel on AI podcast, CEO & Founder Sean Phan joined the team to talk about agentic workflows. He explained how they bridge that human-machine chasm, and articulated some mind-bending visions of the future.
Before we get there, a quick diversion back to earth (or silicon), to unpack agentic workflows.
What are Agentic Workflows?
Agentic workflows mimic that human process we outlined earlier, but for LLMs. They allow an LLM to act like a (human) agent would: producing an output, reflecting on it, polishing and tweaking it before submitting to the user. This all happens internally, during the model’s thinking process.
This is transformative for everything: text, design, video, even code. And PixelML’s customers are already finding novel applications for the pre-made workflows and no-code workflow builder that the platform gives them. According to Sean, one of the most popular agentic AI workflows is lead contact enrichment.
For instance, imagine a lead comes through from a website or blog. The organization may have this person’s email, but they don’t have all the information they need to fully personalize their outreach. With an agentic workflow, an AI model finds their LinkedIn, scrapes it, and creates a short summary. This can then be used to craft deeply personalized messaging and increase the chances of engagement.
Content Production in Regulated Industries: a Critical Use Case for PixelML
Marketers working in regulated industries face especially acute challenges when it comes to content production. For example, medical or pharmaceutical companies have to work within strict boundaries defined by regulations. They need to ensure their marketing and advertising content is both factual and fully compliant with these regulations.
With PixelML, these organizations can bring all relevant rules into one place - the same place where their content is produced. This eliminates the risk of needing to return to square one and rewrite content in light of a problem that they discover at the end of the chain.
Building the Future of Content Supply Chain with Intel Technology
Sean foresees agentic workflows becoming more popular in future, as AI models become more accessible and affordable, lowering the barriers to entry. He also believes it’s possible that building agentic flows will become a specific role. Or, it could become a specialized task undertaken by a new kind of AI agency, similar to the way that enterprises roll out systems like SAP.
PixelML is at the forefront of AI innovation in this domain, making maximal use of the compute resources and compute engineering support they have received in the Intel Liftoff program.
"I was really amazed by the way Intel® Liftoff is organized. We get to talk to an engineer twice a week to help us solve the problems we’re tackling. We even got access to a Gaudi 2 within one and a half weeks of asking. The collaboration and support we get in the program is helping us to achieve our vision." - Sean Phan, CEO & Founder of PixelML
Listen to the full podcast to learn more about what’s wrong with content supply chains, and how PixelML is creating a better alternative with next-generation AI technology. And if you are blazing your own trail in AI (there’s a high chance that you are, if you’re reading this), we at Intel® Liftoff would love to hear from you. Apply to the program and find your launchpad.
As a Software Tools Ecosystem Specialist at Intel, I’ve had the privilege of working on the dynamic GenAI initiative. My focus is on driving engagement with the software developer audience. I'm a proud team member of the Intel® Liftoff for Startups and the Developer Engagement, Relations, and Studio team.