Part 3: When Work Breaks - Rethinking Workflows in Lean Government Teams

In Part 3 of our 5-part series on the lifecycle of work, Emily Nava and Dr. Andrew Hutson dive into one of the most critical—and often overlooked—phases: workflows.

What happens when the structure behind your work starts to break down? When staffing shortages disrupt predictable processes, agencies are often left reacting instead of operating. In this episode, we explore how inconsistent workflows lead to missed tasks, lack of accountability, and reactive fire drills—and what can be done to fix it.

From redefining how work moves across teams to designing workflows that adapt to today’s staffing realities, this conversation focuses on building resilience into the system itself. Because when resources are limited, structure isn’t optional—it’s essential.

Tune in to learn how agencies can move from chaos to clarity by rethinking workflows as the backbone of scalable, mission-driven work.

About "The Interconnectedness of Things"
Welcome to "The Interconnectedness of Things," where hosts Dr. Andrew Hutson and Emily Nava explore the ever-evolving landscape of technology, innovation, and how these forces shape our world. Each episode dives deep into the critical topics of enterprise solutions, AI, document management, and more, offering insights and practical advice for businesses and tech enthusiasts alike.

Brought to you by QFlow Systems
QFlow helps manage your documents in a secure and organized way. It works with your existing software to make it easy for you to find all your documents in one place. Discover how QFlow can transform your organization at qflow.com

Follow Us!
Andrew Hutson - LinkedIn
Emily Nava - LinkedIn 

Intro and Outro music provided by Marser. 

Transcript

WEBVTT

00:00:10.917 --> 00:00:45.476
<v Nava Emily>Welcome back to the interconnectedness of things. I'm your host, Emily Nava, and this is episode three of our five part series, where we unpack the life cycle of work more specifically work in government agencies, Because when teams are stretched thin, the only way to scale responsibly is to structure how work moves. And that means standardizing from the very beginning at intake. All the way through digitization, which is the second phase. And now at the third phase workflows, I'm joined today by my co host Doctor.

00:00:45.477 --> 00:00:49.615
<v Nava Emily>Andrew Hudson. I have a question for you, Hudson.

00:00:49.616 --> 00:00:53.631
<v Hutson>Lay it on me.

00:00:53.632 --> 00:01:08.370
<v Nava Emily>Do you adapt workflows when you don't have the staff numbers you had before?

00:01:08.371 --> 00:01:13.636
<v Hutson>This is a common question coming up with a lot of our customers, isn't it?

00:01:13.637 --> 00:01:19.995
<v Nava Emily>Yep. Seems like the staff they had before is not the staff they have today.

00:01:19.996 --> 00:01:30.236
<v Hutson>Mhmm. Mhmm. And so it really calls into question the the application and design of workflows.

00:01:30.237 --> 00:01:31.035
<v Nava Emily>Mhmm.

00:01:31.036 --> 00:01:52.900
<v Hutson>So $1 to pesos, you probably don't have a predictable series of steps that some work needs to travel along over time. Across various parties to complete in a predicted state. Right?

00:01:52.901 --> 00:01:53.055
<v Nava Emily>Right.

00:01:53.056 --> 00:02:05.516
<v Hutson>What you have is stuff gets assigned. People do the best they can. You don't hear about it until someone complains and then realize nothing was done. And then you go, oh, no.

00:02:05.517 --> 00:02:07.740
<v Nava Emily>What happened? Where did it break down?

00:02:07.741 --> 00:02:41.775
<v Hutson>We're gonna break down. How do we fix this? So Folks are now looking at those situations where there's just a black box of work getting done that they want more insight into. But then also, they want rudimentary steps handled by the computer to free up time for the folks who remain to get done what they want.

00:02:41.776 --> 00:02:53.416
<v Hutson>I was asking on a call today with some folks and confirming that the mission doesn't change just because there was, deferred retirement.

00:02:53.417 --> 00:02:57.816
<v Nava Emily>Right. Same amount of work.

00:02:54.591 --> 00:03:13.111
<v Hutson>And same amount of work, fewer people. What do you do? Well, hire again, which likely that's not gonna be a, first starter, or two, how can I leverage tools? To reduce my involvement in the work.

00:03:13.112 --> 00:03:15.095
<v Nava Emily>Mhmm.

00:03:15.096 --> 00:03:45.435
<v Hutson>Which means you can keep the same scope of work with fewer people because of the use of technology. But it would also mean that if more people came on, that would be a linear growth, if not a multiplier of, of production. Each person could take on, which is always good and and what's important. And then that final thing is, not all of this work is busy work.

00:03:45.436 --> 00:04:14.295
<v Hutson>It affects real people, affects their lives, affects their livelihood. And so the more effective and efficient you can be with the work that needs to be done, the more you're showing up in the right way. For American citizens, which is a big core part of every federal worker I've met is their devotion to serving them citizens.

00:04:14.296 --> 00:04:15.470
<v Nava Emily>Right. Mhmm.

00:04:15.471 --> 00:04:28.000
<v Hutson>They they take what they do seriously. So how can I show up in a same way to take it seriously? Well, the first is take AI out of the equation just for now.

00:04:28.001 --> 00:04:28.275
<v Nava Emily>But wait.

00:04:28.276 --> 00:04:29.040
<v Hutson>How

00:04:29.041 --> 00:04:33.754
<v Nava Emily>AI is the magic button. Right? Everyone needs AI.

00:04:33.755 --> 00:04:42.336
<v Hutson>you know, even the magic beans from Jack and the Beanstalk had a negative impact. So, you know, it's kinda we're gonna be smart about it. You know?

00:04:42.337 --> 00:04:44.450
<v Nava Emily>No. Okay. But we're coming back to it.

00:04:44.451 --> 00:05:16.889
<v Hutson>We'll come back to it. Yeah. We're not saying no. We're just saying, you know, not yet. Most workflows are linear. Sequential. We call them simple sequential here internally, and, one person is typically involved. If anybody has been part of a team, workflows like that immediately fall into a bottleneck because they don't account for different states.

00:05:16.890 --> 00:05:24.336
<v Hutson>And the number of people that could be involved Aren't connected to the workflow.

00:05:24.337 --> 00:05:28.555
<v Nava Emily>And so if that one person is out or retired,

00:05:28.556 --> 00:05:30.736
<v Hutson>Yep. Mhmm.

00:05:30.737 --> 00:05:32.955
<v Nava Emily>then that workflow stops.

00:05:32.956 --> 00:05:40.560
<v Hutson>Or or even they initiated it, and the workflow stalled because it didn't get the next step to proceed.

00:05:40.561 --> 00:05:41.890
<v Nava Emily>Mhmm.

00:05:41.891 --> 00:06:15.536
<v Hutson>Like, that stinks. Right? Another bottleneck. So taking a more sophisticated approach, the best thing is to think about a state machine workflow where the work can move across different states and follow a stage process where it can go back and forward within that state until everything is properly addressed, and then the work is done.

00:06:15.537 --> 00:06:16.816
<v Nava Emily>When you say state,

00:06:15.616 --> 00:06:17.536
<v Hutson>This is different. Mhmm.

00:06:17.537 --> 00:06:23.536
<v Nava Emily>an example of that could be in review or in progress.

00:06:23.056 --> 00:06:24.336
<v Hutson>Exactly.

00:06:24.337 --> 00:06:24.976
<v Nava Emily>Is that what you mean?

00:06:24.496 --> 00:06:34.089
<v Hutson>So think about approvals. Right? You're in a draft. You're in a review. You're in a finalized. Those are different states. And the the document or documents

00:06:32.832 --> 00:06:34.810
<v Nava Emily>Okay.

00:06:34.811 --> 00:07:01.076
<v Hutson>or simply the the task or metadata, reflects what stage it's in. And can move the item back and forth between those stages in a more sophisticated way, which is great, but still can create bottlenecks with one person involved. Okay. So next stage from there is each state has a different person.

00:07:01.077 --> 00:07:01.530
<v Nava Emily>Mhmm.

00:07:01.531 --> 00:07:35.460
<v Hutson>Well, that helps out because it spreads the work a little bit. But maybe each state then gets its own bottleneck. So the the pinnacle in our view is what we built with our swim lane business process notation. So what that means is, rather than assigning a person to a set of tasks, a role is assigned and many people can

00:07:33.997 --> 00:07:35.700
<v Nava Emily>Who?

00:07:35.701 --> 00:07:47.131
<v Hutson>be added to that role to serve to eliminate the assignment bottleneck. Then if you have the different states, well, that eliminates the simple sequential bottleneck.

00:07:47.132 --> 00:07:47.395
<v Nava Emily>Mhmm.

00:07:47.396 --> 00:08:18.595
<v Hutson>And if each role has multiple people, then there's no state or stage that can get a bottleneck either, because you have roles that are assigned to it that can be distributed. What that means in practice sometimes is got a five person team and it used to be guardrails up for each of the roles. It's going to collapse and every person can perform every role as needed. It doesn't work for every workflow that's out there.

00:08:18.596 --> 00:08:42.571
<v Hutson>There might be some strict guidelines, but it certainly opens the door for productivity, as long as you record what role that person is performing so you know to what they were assigned and and for auditing and tracking purposes. So all that is sans AI.

00:08:42.572 --> 00:08:47.034
<v Nava Emily>Yep. And a lot of times, a manager is doing this manually.

00:08:47.035 --> 00:09:23.990
<v Hutson>Mhmm. So it's great to kinda progress through those levels of maturity. Nothing wrong with starting with simple sequential, but you'll quickly find out it's limited. And so then you go to state machine. And then you'll find, oh, I'm still finding bottlenecks. So then you mature up to state machine plus role based assignment and realize, oh, this is a great state. Okay. So you've gone through that evolution with your workflow, or maybe you were more daring and said, let's get let's get to the end of this chapter.

00:09:23.991 --> 00:09:33.355
<v Hutson>K? And now you're looking at the burden of each of the steps. Do you think you do then?

00:09:32.257 --> 00:09:33.675
<v Nava Emily>Mhmm.

00:09:33.676 --> 00:09:42.880
<v Hutson>So you've you've got it removed the bottlenecks, but man, it's still too much volume coming in.

00:09:42.881 --> 00:09:46.000
<v Nava Emily>You evaluate the capacity of my team, of your team.

00:09:45.981 --> 00:09:52.196
<v Hutson>Mhmm. So what if you can't change the volume?

00:09:52.197 --> 00:09:53.740
<v Nava Emily>Spread the work around.

00:09:53.741 --> 00:09:55.340
<v Hutson>Can't spread it around anymore because everybody's

00:09:54.037 --> 00:09:55.900
<v Nava Emily>That's what she can.

00:09:55.901 --> 00:10:00.916
<v Hutson>all already at capacity. The work keeps coming. Now it gets hard. Right? Like,

00:10:00.917 --> 00:10:02.755
<v Nava Emily>Yep.

00:10:02.756 --> 00:10:04.994
<v Hutson>like, I'm no mathematician.

00:10:04.551 --> 00:10:06.435
<v Nava Emily>I guess you just turned out.

00:10:06.436 --> 00:10:11.831
<v Hutson>Yeah. You can burn out. That's one option. You can just let things sit. Option two.

00:10:11.832 --> 00:10:17.150
<v Nava Emily>Mhmm.

00:10:17.151 --> 00:10:24.275
<v Hutson>I don't think either of those are satisfactory to anybody because the mission Yeah.

00:10:21.717 --> 00:10:24.435
<v Nava Emily>No. And does not right.

00:10:24.436 --> 00:10:45.651
<v Hutson>The mission doesn't change. Right? So you're failing the mission at that point. Oh, no. Now you get, you get fatigue and you get depression and you get self questioning and, and then it just starts to spiral down this water drain of despair.

00:10:45.652 --> 00:10:48.180
<v Nava Emily>And that's the episode, everyone.

00:10:48.181 --> 00:10:59.494
<v Hutson>And scene. Okay. In reality, if that happens, it's not good. So something's gotta give.

00:10:58.496 --> 00:11:00.454
<v Nava Emily>Yeah.

00:11:00.455 --> 00:11:05.591
<v Hutson>And I'm the mathematician, but, like, when the numbers don't add up, you gotta do something else.

00:11:05.592 --> 00:11:06.450
<v Nava Emily>Yep.

00:11:06.451 --> 00:11:11.750
<v Hutson>Now who was teaching me that earlier today, she, she ran the numbers.

00:11:11.751 --> 00:11:23.695
<v Nava Emily>Yeah. When you look at events on your calendar and you've got Two of the same event. You do the math, and then you're like, we don't need the same event. Twice, you get rid of one event. That's what the math says.

00:11:23.696 --> 00:11:25.615
<v Hutson>That's business math for you right there. That's MBA.

00:11:25.397 --> 00:11:29.534
<v Nava Emily>Yep.

00:11:29.535 --> 00:11:33.331
<v Hutson>So, Genta can we pull it back in?

00:11:33.332 --> 00:11:34.175
<v Nava Emily>Yep. Please.

00:11:34.176 --> 00:12:18.755
<v Hutson>Okay. So this is the right time when you done this level of maturity in my view to start to entertain a step in your workflow, not an entire workflow, but a step in the workflow. That can be replaced with an agent. An example of this could be, well, if you hook up what everybody's assigned to and you realize that something can't be assigned because of capacity, because there's some parameters that you put in, the AI can then say, well, these are the options I have to respond just like a manager would.

00:12:18.756 --> 00:12:53.995
<v Hutson>It can make the determination of what the next steps does this go back into a queue? Does it get rejected wholeheartedly? Does it get sent to a different team? All these options you can articulate that normally someone would be doing themselves, but now the agent with these fixed parameters is, has a high likelihood of reliability to route these things correctly. Thus, freeing up the person who had to stop with the volume, and trying to make the determination themselves.

00:12:53.996 --> 00:13:05.737
<v Hutson>So there there's other applications. I'm just trying to pick a simple one since I went through a whole explanation of different types of workflows.

00:13:02.437 --> 00:13:22.195
<v Nava Emily>Yeah. And I think that touches on an important part too, of where, where can I implement AI in what step On-It automating the whole workflow or having an expectation that AI is going to be in every step of every workflow?

00:13:22.196 --> 00:13:37.430
<v Hutson>Oh, yeah. I can be proven wrong on this easily given the right circumstances and the right type of work, I'm sure that what I'm saying,

00:13:35.717 --> 00:13:39.750
<v Nava Emily>Mhmm.

00:13:39.751 --> 00:13:52.371
<v Hutson>can be refuted. What my recommendation here on using it as a single step to enhance over time is practical.

00:13:52.372 --> 00:13:54.935
<v Nava Emily>Mhmm.

00:13:54.936 --> 00:14:16.535
<v Hutson>I emphasize more than anything reliability and output more than eliminating people from the process. I have been a strong proponent of using AI to enhance work, to enhance the human.

00:14:16.037 --> 00:14:18.375
<v Nava Emily>Yep.

00:14:18.376 --> 00:14:36.291
<v Hutson>What I have difficulty even today with the enhancements and models is saying, let that sucker run on its own. It needs inspection. It needs guardrails. And maybe that's just right now.

00:14:36.292 --> 00:14:37.731
<v Nava Emily>Well, I think especially in terms

00:14:36.616 --> 00:14:38.131
<v Hutson>Knows what

00:14:38.132 --> 00:14:40.695
<v Nava Emily>of of government work, security is such a

00:14:40.696 --> 00:15:00.390
<v Hutson>the future Well,

00:14:42.637 --> 00:15:01.030
<v Nava Emily>A cornerstone in using AI. And so I think that it's right to expect that AI can be sprinkled throughout a workflow because you don't want it all automated, and you still need your people to be there.

00:15:01.031 --> 00:15:16.876
<v Hutson>that's another great point. Like, One thing I will compliment the agencies that we've worked with directly is its measured approach to adopting large language models.

00:15:16.877 --> 00:15:17.640
<v Nava Emily>Mhmm.

00:15:17.641 --> 00:15:38.855
<v Hutson>They're selective. They're deliberate. I don't think it's out of fear. I think it's out of standards. And that's another way you can be a good steward of that process is Hey, listen. I took this as far as I can with people. If we could replace x,

00:15:37.672 --> 00:15:39.095
<v Nava Emily>Nope.

00:15:39.096 --> 00:16:11.015
<v Hutson>y, and z steps with an agent. Let's start with one and evaluate the results. Okay. It's been pretty consistent. So now I'm gonna stop monitoring it every day and move to a week, and you can start to spread out from there, knowing it can be wrong. Just like if you assign a human to it, but it does free up some time from humans in the process,

00:16:11.016 --> 00:16:11.394
<v Nava Emily>Yep.

00:16:11.395 --> 00:16:15.991
<v Hutson>which to me sounds like a great application. Like, I'd set up for that.

00:16:15.992 --> 00:16:23.511
<v Nava Emily>I agree. And it's lessening the burden of your staff and and not sacrificing quality.

00:16:21.581 --> 00:16:52.175
<v Hutson>Right. And in a world where you have the same amount of work with fewer people, Where you find that time. The vectors that we talked about in the last episode, that helps with this. Helps you get similar products based on the prompt that you give it and the stage in your workflow. And the structuring of that intake helps out with this too.

00:16:52.176 --> 00:17:31.410
<v Hutson>With or without the AgenTic AI, workflow benefits from predictable structure. You can do an unattended trigger to a workflow based off of something that comes into the system matching certain criteria. Oh, I can trigger automatically. That's not AI. That's just a rule that you set up. Well, that's cool for kicking off workflows easily. So a lot of this can benefit with or without any type of AI added in.

00:17:31.411 --> 00:17:36.336
<v Hutson>And when you do add it in, in my opinion, it should be delivered and targeted.

00:17:36.337 --> 00:18:08.616
<v Nava Emily>I agree. I agree with that. Okay. So, again, I'm gonna ask the same kind of question I have been the last three episodes. If you haven't listened to the last two episodes, I highly recommend. We're starting from the first stage of work, which is intake. Then we moved into how you organize your documents. And now we're And how do you keep that work moving through workflows?

00:18:08.617 --> 00:18:32.474
<v Nava Emily>So I'm an agency leader. I've got my standards. I got intake all figured out. I've got my documents organized. Which what questions should I be asking about my current processes or workflows if if I'm if I'm left with less staff than I had a year ago?

00:18:32.475 --> 00:18:50.050
<v Hutson>Mhmm. Well, where a lot of folks are coming to me directly now is they're realizing the bottlenecks. They may not use that terminology, but they're realizing, hey, this has taken way longer than it should, or it's going

00:18:49.097 --> 00:18:50.770
<v Nava Emily>We're installing.

00:18:50.771 --> 00:18:57.655
<v Hutson>slower than We can accept because the work needs to keep on a schedule.

00:18:57.656 --> 00:18:58.715
<v Nava Emily>Mhmm.

00:18:58.716 --> 00:19:09.431
<v Hutson>So when those problem arise, it's a it's an easy trigger to look at workflow. Another part would be I'm I'm losing where the work is.

00:19:09.432 --> 00:19:10.470
<v Nava Emily>Yep.

00:19:10.471 --> 00:19:48.475
<v Hutson>So I want, a way to track that something is out there and a system that can tell me what stage it's in. So I can have some intervention as I see fit. So just greater insights into the work, is aided by workflows. With the I'm trying to think of that example that I was just on. So the the issues have to do with, people don't have the right access to information because nothing gives it to them.

00:19:48.476 --> 00:20:17.196
<v Hutson>Part of the work being assigned. So how do we how do we package up the info that they need and at what stage? There needs to be approvals and something cycles through approvals at at another stage. And then we need a knowledge base that can inform past or future decisions that are extremely similar to Past decisions.

00:20:17.197 --> 00:20:18.316
<v Nava Emily>That's where vector embeddings

00:20:17.376 --> 00:20:18.956
<v Hutson>So then,

00:20:18.957 --> 00:20:19.935
<v Nava Emily>come in.

00:20:19.936 --> 00:20:49.456
<v Hutson>boom. So you circle that all back around to, I have a structured way of getting the data in. I have the ability to OCR the documents. I can create vectors. Now I can leverage that as part of a workflow to say, Hey, I've done this before. Pretty similar. I'm going to take the same approach this time to have more predictability in what we approve. All that stuff starts to fit together, and your investment starts to pay off.

00:20:49.457 --> 00:20:54.440
<v Nava Emily>And your staff is happier because they're not doing five jobs at once.

00:20:54.441 --> 00:21:02.900
<v Hutson>Oh, man. Yeah, of course. Can't believe I didn't say that. Oh, God. You're on this morale.

00:21:02.901 --> 00:21:18.695
<v Nava Emily>Yes. That's what it's all about. Maintaining morale, making work less burdensome and thoughtfully and strategically involving AI, where it makes sense.

00:21:18.696 --> 00:21:20.336
<v Hutson>Approved.

00:21:20.337 --> 00:21:28.310
<v Nava Emily>Herbert. Any last thoughts on workflows before we move on to the next phase?

00:21:28.311 --> 00:21:30.976
<v Hutson>Yeah. Do more of them.

00:21:30.977 --> 00:21:31.376
<v Nava Emily>Few more workflows.

00:21:31.351 --> 00:21:42.780
<v Hutson>Pay yourself back. Yeah. Use workflows as much as possible. And I like this phrase of pay your future self back

00:21:42.781 --> 00:21:44.594
<v Nava Emily>Yep. Mhmm.

00:21:44.595 --> 00:21:59.796
<v Hutson>and you start automating stuff like you're, you should be amazed. Oh my gosh. I know everything that's going on. I know everything is where it should be. I'm able to intervene if something is anomalous.

00:21:59.797 --> 00:22:00.695
<v Nava Emily>Mhmm.

00:22:00.696 --> 00:22:04.436
<v Hutson>And I Roam, I'm able to do that in like a couple of minutes.

00:22:04.437 --> 00:22:15.310
<v Nava Emily>And it's not living in my email inbox, or it's not living on this twenty year old spreadsheet that keeps being added

00:22:15.311 --> 00:22:17.911
<v Hutson>Now everybody knows about it, and I can go to the beach.

00:22:17.912 --> 00:22:34.356
<v Nava Emily>on to. Yep. And that that's really that's really it. Is your staff is free to take some time off. Or if yep.

00:22:29.276 --> 00:22:35.316
<v Hutson>They're happy. They can get on the boat and see a horizon for once,

00:22:35.317 --> 00:22:36.196
<v Nava Emily>And the work doesn't stop.

00:22:35.356 --> 00:22:36.595
<v Hutson>you know,

00:22:36.596 --> 00:22:37.076
<v Nava Emily>Just keeps going.

00:22:37.035 --> 00:22:41.821
<v Hutson>and and no one has to stress that the work is stopping.

00:22:41.822 --> 00:23:12.490
<v Nava Emily>Right. Alrighty. Well, thanks for chatting with me about workflows here. Looking forward to our next episode where we, talk about once the workflow is finished. Record keeping, tracking, auditability, tracking the performance of just workflows in general and how AI can, help or hinder that phase of the workflow.

00:23:12.491 --> 00:23:20.891
<v Nava Emily>So if you found this episode to be helpful, please share it with your network, and we look forward to seeing you in the next one. Thanks.