AI Agent Sprawl: Govern the Lifecycle Before You Backtrack
Episode Summary
Anyone can build an AI agent now, and that is the problem. Tim Crawford, founder and CIO Strategic Advisor at AVOA, coined the term AI agent sprawl. Low-code tools have democratized agent building, so an average user can stand one up in minutes where that once took a pro developer and a team. He names the upside first: agents reach the whole employee base, not only IT. The downside is that people build agents, forget about agents, and move on to the next, and the forgotten ones keep consuming resources until consumption passes the value returned. His answer is governance decided before the agents exist: how each is evaluated, prioritized, managed and sunsetted, because the alternative is backtracking. He drops the premise of a single model and asks for the right smallest number of policies, then separates agents from chatbots and ends on a caution that the path is not all roses.
Key takeaways
The vendor landscape is his starting point rather than his argument. Salesforce Agentforce, Gemini from Google and Copilot from Microsoft are the examples he reaches for first, and he says there was a huge uptick in interest in agents and agentic AI since they last spoke
He extends the list to the larger platform vendors and then states the point himself. After naming Q from Amazon, watsonx from IBM and Joule from SAP, he says everybody is looking at how they can leverage agents into their solutions to increase the value for customers
The caution he attaches to that acceleration is AI washing, which he says there is still a phenomenal amount of. His example is a toaster now improved with AI, and the work he wants done is distinguishing what is agentic from what is not
AI agent sprawl is his own term, and he had just published on it. He says it is becoming an issue for organizations starting down that path, and that he penned a blog post on this very issue earlier that month
The change he identifies is who gets to build. Low-code and no-code tools, GUIs and drag and drop have put very low hurdles in front of building powerful agents, which he describes as democratizing access to AI agents
He names the upside before the downside, and does not treat the trend as a mistake. The gain is exposing this technology and this power to your entire employee base rather than restricting it to IT
What he says is missing is governance shaped around the whole life of an agent. Agents need to be evaluated, prioritized, managed and sunsetted through their life cycle, and it is the absence of that, not the building, that turns volume into sprawl
His description of how sprawl actually accumulates is behavioral rather than technical: everybody has their own set of agents, they build agents and forget about agents, or stop using them and move on to the next agent
The cost is that abandonment is not free. Those agents are still consuming resources: minimal in some cases, he says, and in others it could be something significant. The end state he names is more resources consumed than value being returned
His prescription is about sequence, not tooling. Because demand for building agents will keep rising as the agentic era arrives, the thinking has to happen upfront, so that nobody has to backtrack and work out how to back away from it later
On governance his first instruction is to abandon a premise: you are not going to have just one model, and it is not the same problem as a data warehouse with one common schema
What replaces the single schema is a fluid workspace where data arrives in different shapes, sizes, volumes and velocities and is protected in different ways, and layering state, federal and sovereignty requirements on top gets complicated fast enough that humans cannot effectively manage it
The number he uses to make that concrete is that there are 20 different state privacy laws on the books in the US alone, and that some of them conflict with one another
His resolution is a count rather than a document. You cannot come up with just one, he says, and the target is the right smallest number of policies needed to manage the different work streams and data streams being exposed to agents
The myth he names as probably the biggest is that an agent and a chatbot are the same thing. The chat interface may look the same, but a chatbot requires you to think ahead and build the set of questions and the answers those questions tie to
An agent is set up to do one very specific task, and he is blunt that this makes a single agent limited: an individual agent is not super useful. The power arrives with an orchestration layer that calls other agents, each also doing one specific thing
Working through what it would take to reroute a package, he puts the number at eight or 10 different agents. He counts one doing nothing but validating that Tim Crawford is Tim Crawford and that there is a package, and then immediately corrects himself, calling that probably the second agent
Culture is the first of the three changes he expects within one to three years, and he calls it one of the leading challenges people have in working with agents today
The survey he runs makes culture a measurable barrier rather than an impression. He asked the CIO Think Tank he leads about the biggest challenges with AI agents, and internal culture came back second highest
The top challenge was data: location, strategy and access. He adds that a separate CIO Think Tank survey found AI to be one of the leading factors driving organizations to rethink their data strategy
The second change is greater comfort with automation, because moving into agentic means agents calling other agents and automating some of these processes, and the third is the concept he calls digital agents
The illustration he gives for digital agents is a demo he attributes to Salesforce, at one of their events the previous year. Someone called in and interacted with what sounded like a person, and it was actually a digital agent, which was able to order a product and make some changes to it
He closes on the thorns rather than the roses. There are some along the way that have to be navigated carefully, and what CIOs are contending with today is which steps to take to embrace the technology while staying focused on the value opportunities that lead toward business objectives, without introducing undue risk and governance problems into the fold
About Tim Crawford
Tim Crawford is the founder and CIO Strategic Advisor at AVOA, which he started in 2009 and through which he advises Global 2000 executives and boards on technology strategy. He came to it from more than thirty years inside enterprise IT, including Director of IT Operations at the Stanford Graduate School of Business, CIO and Vice President of Information Technology at All Covered, a division of Konica Minolta, and IT management at Philips Semiconductors. He founded and leads the CIO Think Tank, a peer group of forward-thinking CIOs whose survey findings he cites in this conversation, and he hosts the CIO In The Know and CXO In The Know podcasts. He writes on agentic AI at AVOA, where he published AI Agent Sprawl: Managing Opportunity and Risk in the Enterprise in March 2025. The term is his own. He spoke on the Overcoming Operations and Infrastructure Challenges When Deploying AI to Production panel at the AI Realized Summit in San Francisco.
In this episode
| 00:42 | Welcome, and why David is hosting this week |
| 01:07 | Tim Crawford’s background, AVOA, and the two podcasts he hosts |
| 02:25 | A year of uptick in agents, and the first wave of vendor tools |
| 02:56 | Q, watsonx and Joule: the platform vendors all move at once |
| 03:24 | AI washing, and the toaster now improved with AI |
| 04:06 | The first signs that an organization has AI agent sprawl |
| 04:30 | Where the sprawl problem comes from, and the blog post behind it |
| 04:55 | Low-code, no-code and drag and drop democratize agent building |
| 05:36 | An average user can build an agent in a matter of minutes |
| 05:36 | The missing piece: evaluated, prioritized, managed, sunsetted |
| 05:36 | Build agents, forget agents, move on to the next agent |
| 06:35 | Forgotten agents still consume resources |
| 06:57 | The overhead: policy, regulatory, compliance, security, data privacy |
| 07:24 | Think about it upfront, or backtrack out of it later |
| 07:56 | What governance frameworks multiple agents need |
| 08:16 | You are not going to have just one model |
| 08:35 | A fluid workspace: shapes, sizes, volumes, velocities, sovereignty |
| 09:02 | 20 state privacy laws in the US, some conflicting with each other |
| 09:02 | Not one policy, the right smallest number of policies |
| 10:07 | The biggest myth: that an agent and a chatbot are the same thing |
| 10:07 | What a chatbot requires you to build ahead of time |
| 11:03 | One agent, one task, and why a single agent is not super useful |
| 11:03 | The orchestration layer, and where the power actually comes from |
| 11:41 | Rerouting a package, worked through step by step |
| 12:00 | Eight or 10 agents for one request |
| 12:40 | Why the orchestration layer is what makes agentic powerful |
| 13:06 | A tectonic degree of sophistication that chatbots do not have |
| 13:26 | Where humans and AI agents are heading in one to three years |
| 13:52 | Cultural acceptance as the first thing to change |
| 14:19 | The CIO Think Tank survey: internal culture is the second challenge |
| 14:46 | Data location, strategy and access is the first |
| 15:34 | Greater comfort with automation as agents call other agents |
| 15:57 | Digital agents, and the Salesforce demo that sounded like a person |
| 16:27 | Not a path that is just all roses |
| 16:48 | The thorns along the way, and what CIOs are contending with today |
| 18:11 | Sending an agent as the guest next time |
In Tim’s words
“an average user can build an agent within a matter of minutes”
Tim Crawford (05:36)
“everybody’s got their set of agents and they build agents and they forget about agents, or they stop using agents and they’re onto the next agent”
Tim Crawford (05:36)
“you’re gonna end up with more resources consumed than what you’re actually getting value from, and that inherently is why sprawl becomes a problem”
Tim Crawford (06:35)
“you need to be thinking about that upfront so that you don’t end up with this situation and then have to try and backtrack and figure out how to back away from it”
Tim Crawford (07:24)
“you’re not gonna have just one model, so just put it out of your head”
Tim Crawford (08:16)
“You can’t come up with just one policy.”
Tim Crawford (09:02)
“an individual agent is not super useful. It is useful, but it’s not super useful.”
Tim Crawford (11:03)
“There’s a complete tectonic degree of sophistication that comes with agents that you don’t have with chatbots.”
Tim Crawford (13:06)
“this is not a situation of, great, it’s a path and it’s just all roses”
Tim Crawford (16:27)
Tim Crawford
Tim Crawford on LinkedIn: Where he posts, and the best place to follow his work with CIOs on agentic AI and technology strategy
AVOA: His firm, where he is founder and CIO Strategic Advisor, advising Global 2000 executives and boards on technology strategy
Tim Crawford’s writing at AVOA: His author archive, where the agentic AI posts referenced in this conversation sit alongside his ongoing CIO research
The two posts he refers to on air
AI Agent Sprawl: Managing Opportunity and Risk in the Enterprise: The post he points to at 04:30 as the one he had just published on this issue. It sets out the lifecycle argument he makes here, that agents should be managed with a lifecycle approach rather than a project mentality, with policies covering training, monitoring and decommissioning
Welcome to the Agentic Era and the rise of humanistic interactions: The second post, which he refers to at 07:24 as another blog post he has written on the agentic era. It makes the same case for agents as an advance on chatbots that he draws out at 10:07
Ideas and terms discussed
AI agent sprawl: His own coinage, and the spine of the episode. Building an agent is now cheap enough that anyone can do it, so agents accumulate faster than anyone tracks them. People build them, forget them, or stop using them and move to the next, and the abandoned ones keep drawing resources until consumption exceeds the value being returned
The agent lifecycle: What he says is missing when sprawl sets in, named as four stages: evaluated, prioritized, managed, sunsetted. His point is that the whole sequence has to be decided before the agents exist rather than after, and that the absence of that governance, not the building of agents, is what turns volume into sprawl
Democratizing access to AI agents: His description of the change that causes the problem, and he frames it as a gain first. Low-code and no-code tools, GUIs and drag and drop have replaced what once required a pro developer, a whole team and a lot of time, which means the capability reaches the entire employee base instead of being restricted to IT
AI washing: The term he uses for the marketing layer over the real shift, which he says there is still a phenomenal amount of. His illustration is a toaster now improved with AI, and the discipline he asks for is distinguishing what is genuinely agentic from what is not
The right smallest number of policies: His formulation for AI governance, positioned against the obvious wrong answer. One policy cannot work, because data arrives in different shapes, sizes, volumes and velocities and is protected in different ways, with state, federal and sovereignty requirements layered on top. He notes that some of the 20 state privacy laws conflict with one another. What he asks for instead is the right smallest number of policies, enough to cover the actual work streams and data streams being exposed to agents and no more
Agents are not chatbots: The one he calls probably the biggest myth, and later probably the most common misnomer. A chatbot requires you to think ahead and build the set of questions along with the answers they tie to. An agent is set up to do one specific task and does not require that anticipation, and what he calls a tectonic degree of sophistication sits behind an interface that looks similar
The orchestration layer: His answer to why one agent is not enough. A single agent is useful but not super useful, because it does one specific thing. Power comes from an agent that ties into an orchestration layer that calls other single-purpose agents, which is what turns eight or 10 narrow agents into one completed request
Digital agents: The third change he expects within one to three years: an agent that acts a lot like a human rather than one that answers like a system. He separates this from cultural acceptance and from comfort with automation, which are the first two
Named on air
Salesforce Agentforce: One of his first examples at 02:25 of the agent tools that arrived over the preceding year. Salesforce returns at 15:57 for a different reason, as the company he attributes the digital agent demo to
Gemini and Copilot: Named at 02:25, Gemini from Google and Copilot from Microsoft, in the same list of tools that came to market as interest in agents rose
Amazon Q, IBM watsonx and SAP Joule: His examples at 02:56 of the larger platform vendors moving at the same time, offered as evidence that everybody is looking at how to leverage agents into their solutions
Samsung: Raised by David at 03:53, not by Tim, as the example of AI features appearing in every product down to refrigerators and microwaves. David qualifies it in the same breath, that the AI-powered features in those products are real, and at 04:06 turns it into the point that every company is now an AI company
The CIO Think Tank: The peer group of forward-thinking CIOs he says at 14:19 that he leads, and the source of both survey findings on this page: internal culture as the second highest challenge with AI agents, and AI as a leading factor driving organizations to rethink their data strategy
AI Realized Summit: Named at 00:42 as the event he spoke at the previous fall in San Francisco, and at 01:07 as the Overcoming Operations and Infrastructure Challenges When Deploying AI to Production panel held at UCSF, and again at 17:32 as returning to San Francisco in 2025
CIO In The Know and CXO In The Know: The two podcasts he hosts, named by David at 01:07 and again at 17:32 as a place to go deeper on his work
Frequently Asked Questions
-
AI agent sprawl is what happens when an organization accumulates more AI agents than it tracks or retires, so abandoned agents keep consuming resources without returning value. Tim Crawford of AVOA, who coined the term, traces it to low-code and no-code tools that let an average user build an agent in minutes, which democratizes access. The same capability once required a pro developer, a whole team of folks and a lot of time. The pattern he describes is behavioral: people build agents, forget about them, or stop using them and move on to the next one. The end state is more resources consumed than value being returned, plus the overhead of governing an agent infrastructure nobody planned.
Transcript 04:30 to 07:24
-
AI agent sprawl is caused by easy agent building running ahead of lifecycle governance. Low-code and no-code tools, GUIs and drag-and-drop interfaces are what raise the volume, and Tim Crawford of AVOA treats that part as a gain, because the capability reaches the entire employee base instead of being restricted to IT. What he calls the downside is that nothing then decides how agents are evaluated, prioritized, managed and sunsetted through their life cycle. Without that, agents accumulate faster than anyone retires them.
Transcript 04:55 to 06:57
-
You prevent AI agent sprawl by deciding the agent lifecycle before the agents exist, covering how each one is evaluated, prioritized, managed and sunsetted. Tim Crawford of AVOA argues the timing is the whole point: demand for building agents will keep rising as the agentic era arrives, so the thinking has to happen upfront rather than after the fact, when the work becomes backtracking and figuring out how to back away from it. He says it is the absence of that governance, not the building of agents, that lets forgotten agents keep drawing resources.
Transcript 05:36 to 07:24
-
The difference between an AI agent and a chatbot is that a chatbot requires you to anticipate the questions in advance and build the answers they tie to, while an agent is set up to perform one specific task without that anticipation. Tim Crawford of AVOA says the belief that the two are the same is probably the biggest myth or confusion he encounters, and that the similarity is only at the interface, because both can be reached through a chat window. Behind it he describes a complete tectonic degree of sophistication that chatbots do not have.
Transcript 10:07 to 13:06
-
Multiple AI agents work together through an orchestration layer, where one agent calls other agents that each do a single specific thing. Tim Crawford of AVOA is direct that a single agent is limited on its own, describing an individual agent as useful but not super useful, and locates the power in the layer that coordinates them. His worked example is rerouting a package, which he breaks into eight or 10 agents: one validating that the person is who they say they are, another determining which package is meant, then each subsequent stage in turn, with every agent doing its own piece well and nothing else.
Transcript 11:03 to 12:40
-
Governing AI agents across departments means writing the right smallest number of policies rather than a single policy for everything. Tim Crawford of AVOA says the first premise to drop is that there will be a single model, and that this is not the same problem as a data warehouse with one common schema. What replaces it is a fluid workspace where data arrives in different shapes, sizes, volumes and velocities and is protected differently, and layering state, federal and sovereignty requirements on top gets complicated fast enough that humans cannot manage it effectively. He notes there are 20 different state privacy laws on the books in the US alone, some of which conflict with one another.
Transcript 08:16 to 09:02
-
Data is the biggest challenge CIOs face in adopting AI agents, specifically its location, strategy and access, with internal culture second. Tim Crawford of AVOA surveyed the CIO Think Tank he leads, a peer group of forward-thinking CIOs, and reports those two as the top challenges in that order. He adds that a separate survey of the same group found AI to be one of the leading factors driving organizations to rethink their data strategy, which is what connects the two findings.
Transcript 14:19 to 14:46
-
Over the next one to three years, three things change: people become culturally more willing to work with agents, they grow more comfortable with automation as agents call other agents, and digital agents arrive that act a lot like a human. Tim Crawford of AVOA puts cultural acceptance first because he considers it one of the leading challenges today, and expects it to ease as agents get smarter than the chatbot that cannot understand a question outside its knowledge base. For digital agents his example is a demo he attributes to Salesforce, where someone called in and interacted with what sounded like a person that was actually a digital agent, and the agent ordered a product and made changes to it. He ends on a caution rather than a forecast: the path is not all roses, and there are thorns along the way.
Transcript 13:52 to 16:48
-
AIRealized-Ep17-Tim-Crawford
[00:00:00] Christina Ellwood: AI Realized, the podcast about everything that is new, now, and next for enterprise executives deploying AI. Hosted by myself, Christine Aylward, and my collaborator, David Yakobovitch.
[00:00:17] David Yakobovitch: Our guests will share what's driving AI adoption, use cases, and business models for the data and AI economy
[00:00:42] David Yakobovitch: Welcome back, listeners. I'm David Yakobovitch, one of your hosts for the AI Realized podcast. I'm stepping in this week to talk about enterprise AI deployment, and thrilled to have with you one of our guest speakers from this past fall's AI Realized Summit in San [00:01:00] Francisco. We actually have today the guest is Tim Crawford, who is founder and CIO strategic advisor of AVOA.
[00:01:07] David Yakobovitch: He was on our Overcoming Operations and Infrastructure Challenges When Deploying AI to Production panel that we had at UCSF. And diving more into his background, Tim Crawford is not only the founder of AVOA, he's one of the most influential CIOs in the US. He advises large global enterprises and serves as a podcast host for both the CIO in the Know and CXO in the Know.
[00:01:35] David Yakobovitch: His work focuses on how technology, especially AI, can be used as strategic levers to drive enterprise transformation. Tim, it's great to reconnect with you and have you on the show.
[00:01:48] Tim Crawford: Likewise. Thanks, David, for the invite. I'm looking forward to our conversation.
[00:01:52] David Yakobovitch: Yeah, 100%. And I know last year in 2024, we started talking about big topics of what would [00:02:00] happen in '25, like the rise of AI agents.
[00:02:02] David Yakobovitch: And here we are, as we know, with AI agent technology- ... coming to market. It's evolving rapidly. Everyone's talking about it. So for you, if you could share with our listeners to, to start, what are some of the most promising enterprise use cases you're seeing with AI agents, or just get us started on what you're seeing with the industry.
[00:02:25] Tim Crawford: So we did talk about agents last year, and there was a huge uptick in, in interest around where agents and agen- agentic AI, where it was gonna go, how it was gonna really hit the ground running. And we saw some pretty significant, uh, tools that were coming out, solutions that were coming out. Just to throw some examples out there, Salesforce's Agentforce, or even some of the Copilot and Gemini from Google and Copilot from Microsoft, those were other tools.
[00:02:56] Tim Crawford: But even some of the bigger organizations [00:03:00] like SAP and IBM and Amazon, they all have tools, whether it's Q from Amazon or Watsonx from IBM or Joule from SAP. The point is- Everybody is looking at how they can leverage agents and agentic tech- technology and approaches into their solutions to increase the value for customers.
[00:03:24] Tim Crawford: And that's something that we've seen accelerate pretty demonstrably in the past 12 months, which is incredibly exciting. But it's important to distinguish between what is agentic and what isn't in the light of there's still a phenomenal amount of AI washing that is coming over. "Here's my toaster, now improved with AI," for example.
[00:03:49] Tim Crawford: And so we wanna make sure that we're focused on those kind of value propositions.
[00:03:53] David Yakobovitch: Yeah, for sure. Definitely in the news this year, we've seen how Samsung has turned every product into AI refrigerator, [00:04:00] AI microwave. So it does beg the question. To be frank, though, there are AI-powered features in there, so- Yes, there are
[00:04:06] David Yakobovitch: it's the evolution where not only is every company a tech company, every company is an AI company. Now, speaking to agents, so you've coined the term AI agent sprawl. What are the first signs that an organization is experiencing this problem when they're looking to use agents and agentic systems, and what immediate steps should they take?
[00:04:30] Tim Crawford: That's right. AI agent sprawl is becoming an issue for those that are starting down that path, and I just penned a, a blog post on this earlier this month on this very issue. And just for listeners so that you understand what ag- where AI agent sprawl is coming from, is you think about the use of AI and how it's progressed over time.
[00:04:55] Tim Crawford: There was a fair amount of low-code/no-code type solutions that were using [00:05:00] AI and then to build AI agents. And so you start to use GUIs and drag and drop and really low hurdles to be able to build these incredibly powerful agents, and we're seeing more of that as we go through time. What that is essentially doing is democratizing access to AI agents, whereas historically, if we were to apply how we would traditionally build this kind of really robust technology, it would require a pro developer, a lot of time, a whole team of folks, a lot of time to build this kind of sophistication.
[00:05:36] Tim Crawford: And now what we're seeing is that an average user can build an agent within a matter of minutes. And so what's great about that is you can start to expose this technology and this power and these opportunities to your entire employee base as opposed to restricting it to IT. [00:06:00] The downside of that is you start to run into this sprawl problem, and the sprawl problem is without a good degree of governance on how agents are evaluated, prioritized, managed, sunsetted through their life cycle, without that kind of governance in place, the problem is you run into this situation where everybody's got their set of agents and they build agents and they forget about agents, or they stop using agents and they're onto the next agent.
[00:06:35] Tim Crawford: The problem with that is those agents are still consuming resources. In some cases, it might be minimal. In other cases, it could be something significant The reality is you're gonna end up with more resources consumed than what you're actually getting value from, and that inherently is why sprawl becomes a problem.
[00:06:57] Tim Crawford: And then there's the overhead of [00:07:00] managing this entire infrastructure of agents and agentic technology that you have to consider when you start to think about governance and policies and regulatory and compliance requirements. And then there's the whole security and data privacy and how those governance models get exposed to users and administrators that are building agents and using agents.
[00:07:24] Tim Crawford: So there's a lot of moving parts here, but the sprawl problem is one of we're gonna see a huge uptick in demand for building agents, especially as we truly get into the agentic era, which is another blog post that I've written about. The challenge is you need to be thinking about that upfront so that you don't end up with this situation and then have to try and backtrack and figure out how to back away from it.
[00:07:52] Tim Crawford: And so that's the gist behind the agent sprawl problem.
[00:07:56] David Yakobovitch: It makes a lot of sense. And Tim, in your comments, you [00:08:00] mentioned a couple times about governance, and I imagine organizations and executives are thinking, "What governance frameworks do I need if I'm gonna implement multiple AI agents across departments?"
[00:08:12] David Yakobovitch: Would you have any suggestions for executives getting started here?
[00:08:16] Tim Crawford: When you think about governance, number one, you're not gonna have just one model, so just put it out of your head. You're not gonna have one model. And this isn't the, this isn't the same problem as you have when you start to think about, like, a data warehouse and a common schema that you use.
[00:08:35] Tim Crawford: You're gonna have this very fluid workspace that you're gonna need to work in, and the data comes in all different shapes, sizes, volumes, velocities. It's protected in different ways, and especially as you start to layer state and federal and sovereignty requirements into the mix, it gets pretty complicated pretty fast, to the point that humans actually can't [00:09:00] effectively manage it.
[00:09:02] Tim Crawford: Like, for example, within the US alone, there are 20 different state privacy laws on the books In the US. That's just in the US, 20 states. Oh, and by the way, some of them conflict with one another. So how do you start to manage through that? You can't come up with just one policy. You have to be able to address the right smallest number of policies in order to effectively manage the different work streams and data streams that you're gonna expose into agents.
[00:09:40] Tim Crawford: And that's really what it comes down to.
[00:09:43] David Yakobovitch: And it's not only, as you're mentioning, the different workflows and how to think about governance, but I imagine executives have a lot of misconceptions also here when they're thinking about implementing AI agents. It's so new and it's what is, what is prod [00:10:00] versus staging?
[00:10:00] David Yakobovitch: So do you have any suggestions on these, like, common myths that executives are experiencing?
[00:10:07] Tim Crawford: Probably the biggest myth or confusion, maybe is a better way to put it, is that there's the belief that an agent and a chatbot are one and the same And those two are very different. Now, you might have a similar human interface to an electronic system in the sense that there's a chat interface that you're using, but the difference, the core difference between agents and chatbots is, or one of the core differences, there are many of them, is that with chatbots, you have to think ahead and, okay, if I'm working with a certain user base and I want this interface to be able to address certain problems or answer certain questions, I have to build that set of questions and build the answers that those [00:11:00] questions are going to tie to.
[00:11:03] Tim Crawford: With agents, you're not doing that. With agents, you are setting up a specific agent to do a very specific task. Now, when someone interacts with an agent, one of the things that you quickly realize is an individual agent is not super useful. It is useful, but it's not super useful. Where it really gets powerful is where you have an agent that then ties into an orchestration layer that then calls other agents that also only do one very specific thing.
[00:11:41] Tim Crawford: So think about rerouting a package, for example. Great, so the initial agent is gonna determine that has to go to a different system. We have to figure out where the package is, who has access to the package. We have to validate who the person is. We have to then look at what are the different possibilities.[00:12:00]
[00:12:00] Tim Crawford: Are there costs associated with that? This might be eight or 10 different agents that have to come in, 'cause you have one agent that might be just simply validating that Tim Crawford is Tim Crawford, and that there is a package. Actually, that was probably the second agent of determining, okay, we've validated who Tim Crawford is.
[00:12:19] Tim Crawford: Now let's figure out what package is Tim Crawford talking about. Great, so we've got that. And then getting to the next stage, and getting to the next stage. And so each of those steps are essentially a unique agent They work on that particular part really well, but that's all that they do is just that one piece.
[00:12:40] Tim Crawford: And that's why this orchestration layer, and especially as you start to think about agentic, becomes really powerful. And that's the one piece that I find to be probably the most common misnomer, is that agents are just like chatbots. And while on the surface there may be some visible truth [00:13:00] to that in terms of how the user interacts, behind the scenes, they are vastly different.
[00:13:06] Tim Crawford: There's a complete tectonic degree of sophistication that comes with agents that you don't have with chatbots.
[00:13:14] David Yakobovitch: And so diving a little bit deeper here, y- you've discussed the agentic era both in writing and through our conversation, and you just shared with our listeners, you differentiated between chatbots and agents.
[00:13:26] David Yakobovitch: But if we think forward-looking, how do you see the relationship between humans and these more sophisticated AI agents evolving in the next one to three years?
[00:13:37] Tim Crawford: Yeah, one to three years is actually a lot of time in this era because things are accelerating so quickly. Just look at generative AI. Generative AI is something that's just happened in the last, what, two years?
[00:13:52] Tim Crawford: Little more than two years, and look at how far we've come, and now we're talking about agentic. I think in the next one to three [00:14:00] years, you'll see a couple things happen. One is culturally, you'll see more acceptance of people willing to work with agents. Today, culturally, that's one of the, one of the leading challenges that people have in working with, working with agents.
[00:14:19] Tim Crawford: I have a, a group of forward-thinking CIOs that I work with, and they're part of a CIO think tank that I lead, and, uh, we surveyed them to, to ask them what are some of the biggest challenges that they're seeing with AI agents, and internal culture was the second highest issue with, or challenge with AI agents and being able to adopt AI agents.
[00:14:46] Tim Crawford: The top one, by the way, is data, location, strategy, access. Data strategy is one of the top issues that CIOs are having to work through right now. And by the way, [00:15:00] did another survey with the CIO think tank, and AI was one of the leading, uh, factors that is driving organizations to rethink their data strategy.
[00:15:11] Tim Crawford: So you can see this kinda tie in. So I think in the next one to, one to three years, we'll see a couple things. One is- Culturally, we'll be more accustomed to using agents and engaging with agents because they're gonna be smarter. It's not gonna be the, the dumb chatbot that can't understand my question or doesn't have my situation predefined within its, within its knowledge base.
[00:15:34] Tim Crawford: You don't have to do that. And then the second thing that I think you'll see is greater comfort with regards to automation, 'cause when you start to move into agentic, you're talking about agents calling other agents and automating some of these processes And then the third piece is this concept of digital agents.
[00:15:57] Tim Crawford: And when you get to digital agents, you're [00:16:00] essentially creating an agent that acts a lot like a human. And we saw this with companies like Salesforce did a demo at one of their recent events last year, where they showed calling in and being able to interact with what sounded like a person, but it was actually a digital agent that was able to order a product and make some changes to a product.
[00:16:27] Tim Crawford: And I think we'll see more of that and more comfort with that over the next one to three years. But that's the great opportunity about this. Let's not misunderstand that there are some hurdles we have to get over in order to achieve that. So this is not a situation of, great, it's a path and it's just all roses.
[00:16:48] Tim Crawford: There are some thorns along the way that we have to be careful and have to navigate through as we go through this process. And that's what CIOs are contending with today, is [00:17:00] what does that pathway look like? What are those opportunities, and what are the steps that I need to take for my organization so that I can try and embrace and leverage this technology and this innovation as best as possible, but at the same time, make sure that I'm focused and prioritized around the value opportunities that lead toward my business objectives without introducing undue risk and governance problems into the fold.
[00:17:30] Tim Crawford: And so that's where CIOs are today.
[00:17:32] David Yakobovitch: Wow, you've given us so much knowledge today, Tim, and I can't wait to learn more about that and see that at our AI Realized Summit 2025 that'll be back in San Francisco. And to dive deeper into your work, from things like you mentioned, the CIO Think Tank, to listening and tuning into your episodes of CIO In The Know and CXO In The Know podcasts as part of AVOA.
[00:17:56] David Yakobovitch: So thank you so much for joining us. We've really, uh, loved [00:18:00] deepening this relationship from AI Realized. And who knows? Maybe next time we'll have some AI agents supporting us on one of these sessions.
[00:18:11] Tim Crawford: It... Actually, I'll send my agent to be a guest on your podcast instead of the real me.
[00:18:19] David Yakobovitch: We love it. We'll use the Eleven Labs voice of yourself, so that'll work quite well.
[00:18:25] David Yakobovitch: For all of our listeners tuning in, this has been Tim Crawford of AVOA, founder and CIO strategic advisor here on the AI Realized podcast. Tim, thanks so much for joining us today.
[00:18:36] Tim Crawford: Thank you, David. It's been a pleasure.
[00:18:37] music: AI, where the future is bright
[00:18:56] [00:19:00] Right