Developers are hearing a lot about agentic AI lately, that’s for certain. What’s less certain is what the term actually means and whether it’s A Thing you need to learn about. Is “agentic AI” just a buzzword, or is it the sea change it seems?
There’s no denying that “agentic AI” is a buzz-worthy term. But there is a there there. In this post, we’ll give you a big-picture overview of what AI agents are, what they can do, and how to think about them. This is a good place to start if you’re new to the notion of autonomous AI agents.
When we talk about AI agents or agentic AI, we’re not all talking about the same When we talk about AI agents or agentic AI, we’re not all talking about the same thing. As a recent TechCrunch piece by Maxwell Zeff and Kyle Wiggers put it, No one knows what the hell an AI agent is. Or, at least, not everyone can agree on a definition for what, exactly, an AI agent is or does, even as the industry grows increasingly bullish on the concept. The TechCrunch piece surveys the way titans like OpenAI, Anthropic, Google, and Microsoft describe agentic AI differently, highlighting the term’s plasticity. Much like the term “AI” itself, agentic AI is still coming into focus.
Caveats aside, agentic AI refers to autonomous AI systems that make decisions to achieve specific goals with minimal need for human oversight and intervention. Simply put: Generative AI creates content; agentic AI solves problems on a user’s behalf.
“The agentic AI system understands what the goal or vision of the user is and the context to the problem they are trying to solve,” AI expert Enver Cetin told the Harvard Business Review. The HBR article teases out three key variations between agentic AI and the generative AI methods we’re by now conversant in:
- Agentic AI is targeted on making choices, not producing content material.
- Moderately than responding to human prompts like generative AI methods do, agentic AI methods are set to work towards particular targets like maximizing effectivity, boosting buyer satisfaction, or growing gross sales.
- AI brokers can perform advanced sequences of duties in furtherance of their targets.
To convey this idea right down to earth, right here’s a high-level instance of agentic AI we heard from the stage at TDX final month:
Say you’re in a fender bender. As an alternative of calling roadside help to kick off the lengthy, concerned strategy of getting your automobile towed and repaired, you ping an AI agent who can name emergency companies if essential, contact a towing service, assist you to doc harm to your automobile, floor the related insurance coverage data (like the quantity of your deductible and whether or not your coverage covers a rental automobile), make a listing of the best-reviewed physique outlets in your space and request estimates from them, and coordinate with the insurance coverage firm (or the insurance coverage firm’s autonomous brokers!) all through the entire course of.
The distinction between that set of interconnected actions, all geared towards a selected objective, and the content material a generative AI system produces when prompted with a query a couple of automobile accident tells you a large number in regards to the performance hole between agentic and generative AI. You inform the agent the place you wish to go and it will get you there. If generative AI offers you a map, agentic AI picks you up and takes you.
Due to their capability to orchestrate advanced sequences of occasions and apply cognitive reasoning in furtherance of their targets, agentic AI methods give customers the chance to automate workflows they couldn’t automate earlier than. The post-crash sequence outlined above is only one instance.
Instruments that make it simpler for non-technical folks to create software program are nothing new, in fact. Low-code and no-code instruments have existed for many years. Generative AI-powered coding instruments like Copilot have allowed non-programmers to put in writing a bit of code that will increase their effectivity by automating time-consuming and repetitive duties. However agentic AI methods characterize an exponential step ahead when it comes to how a lot they empower folks with out coding expertise to orchestrate difficult, multi-step processes. And, crucially, they permit folks with coding expertise to do an entire lot extra, too.
AI brokers are usually higher than generative AI fashions at organizing, surfacing, and evaluating information. In concept, this makes them much less liable to hallucinations. From the HBR article: “The better cognitive reasoning of agentic AI methods signifies that they’re much less prone to endure from the so-called hallucinations (or invented data) frequent to generative AI methods. Agentic AI methods even have [a] considerably better potential to sift and differentiate data sources for high quality and reliability, growing the diploma of belief of their choices.”
Regardless of widespread adoption, builders’ lack of trust within the output of generative AI methods persists. If AI brokers are much less liable to hallucinations, may they improve the diploma of religion builders are prepared to position in AI?
From conducting deep analysis into potential clients prematurely of gross sales calls to creating suggestions to enhance course of effectivity to offering threat indicators throughout monetary methods, AI brokers unlock a world of potential use instances that aren’t simply developer-centric. One firm is leveraging agentic AI to help healthcare providers get paid for their work with much less back-and-forth with insurance coverage corporations.
One other set of use cases for agentic AI is the rote-yet-crucial work builders do: testing and reviewing code, writing pull requests, error dealing with, helpdesk operations, risk scanning and safety monitoring, and extra. Agentic AI is poised to transform developer workflows by automating extra of those processes, turning software program growth right into a collaborative course of during which the AI agent executes towards the objective and constraints specified by the human person.
On the Stack Overflow podcast, we interviewed an engineer at Diffblue about how they’re utilizing agentic AI to test complex code at scale. AWS instructed us how they saved 4,500 years of developer time upgrading Java across their humongous codebase (greater than 30,000 packages). We talked with the cofounder and CEO of a startup constructing AI brokers to review code and write pull requests. And an IT director and Salesforce architect we met at TDX is utilizing AI brokers to construct an error dealing with system for his small org.
From our viewpoint, the AI agent doesn’t substitute builders. It frees them to concentrate on what’s sometimes called “higher-order” duties: artistic, strategic, revolutionary or architectural work. It permits them to zero in on the facets of their work the place they’re most passionate or the place they’ll have the most important impression. The perfect agentic AI options will give builders time and vitality again whereas serving to them study new instruments and applied sciences—simply as their most popular generative AI instruments have performed.
Stack Overflow has been a tried-and-true developer useful resource for greater than 15 years, so we’ve seen applied sciences rise and fall, traits come and go. Agentic AI is a paradigm shift on the order of the emergence of LLMs or the shift to SaaS. That’s to say, it’s an actual factor, however we’re not but near understanding precisely the way it will change the way in which we dwell and work simply but.
The adoption curve for agentic AI can have its challenges. There are questions wherever you look: How do you place AI brokers into manufacturing? How do you take a look at and validate code generated by autonomous brokers? How do you cope with safety and compliance? What are the moral implications of counting on AI brokers? As all of us navigate the adoption curve, we’ll do our greatest to assist our group reply these questions. (In spite of everything, answering questions is our complete factor.). Whereas constructing brokers may rapidly develop into simpler, fixing for these downstream impacts remains to be incomplete.
Typically when a brand new method or a brand new piece of tech breaks onto the scene, we develop into fixated on the hype round that tech as a substitute of specializing in its potential to unravel precise issues. Moderately than dashing to undertake agentic AI ASAP, our advice is that builders (and their managers) take a deep breath and ask themselves: What issues can I resolve for my workflows, workforce or employer with AI brokers that I can’t resolve with out them? A transparent reply to that query is the place to begin for contemplating if agent adoption can profit you, and the place it’ll take know-how subsequent.
