The Real-World AI Workflows Used by The Head of Delivery at an Asana Implementation Agency

I stay close to the front lines and manage a couple of key clients at any given time, so I'm also in front of clients weekly.

How AI is Reshaping Project Delivery Roles

At Cirface, we’re spending far less time on small tactical tasks than before. Anything that can be automated in Asana is now handled for us. That shift began even before the AI boom, with Asana’s native rules and automations, but AI has added an entirely new layer.

The tool can now apply reasoning and context, which means it can take on more of the upfront thinking and organizing before I step in as the human user.

You don’t really have to set this up yourself. Traditional Asana rules are binary: trigger, optional condition, and action.

With AI, the logic layer gets smarter out of the box. Instead of “if priority = high, then do X,” you can tell AI what high, medium, or low looks like in natural language, even attach documents like brand guidelines, and then let it decide.

The AI then applies reasoning and context in ways static rules never could, which means it can handle the upfront thinking and pass a more accurate output to the team.

As a result, my focus has shifted. I spend more time designing and refining these AI-powered workflows, while at the same time gaining back a surplus of hours that used to be lost to repetitive execution like tracking metrics.

This frees me up to concentrate on higher strategic value work like improving delivery processes, supporting the team, and making sure projects run more effectively overall.

Same Methods as Before; Just Lighter

So, our methods haven’t really changed, but the way we run them has become lighter.

Here's an example. With Asana, visibility into status and priorities is already built in, so we don’t need daily meetings or constant updates. AI then takes that a step further.

Asana now includes an AI-assisted Project Status update. Picture a shared project where creative and development are all working for a client. Tasks are coming in, comments are piling up, subtasks, dependencies, attachments, and all the usual chaos.

Normally, you’d pick a status (On Track, At Risk, Off Track) and then write a summary: what happened, blockers, next steps, action items, and who needs what. Useful, but if you are juggling ten or more projects, writing those weekly updates is a grind.

Enter "Draft with AI". When you start a status update, Asana can scan the entire project history, including completed tasks, comments, subtasks, dependencies, and attachments. It then generates a first draft that explains in plain English “how things are going” and “what happened last week.” You stay in the loop as the human editor: review the draft, tweak the tone, add any nuance, and post.

It is simple: AI does the heavy lifting, you provide judgment. For anyone managing regular status reports across multiple projects, this can easily save an hour or two each week without sacrificing clarity for your team or your execs.

Lead Smarter in the Age of AI

Join the DPM community for access to exclusive content, practical templates, member-only events, and weekly leadership insights - it’s free to join.

Name

This field is for validation purposes and should be left unchanged.

Name*

First name* Last name*

Business email*

This field is hidden when viewing the form

Free Account Module

Bottom BarLanding PageFooterInlinePop UpSidebarSSO

LinkedIn profile

Seniority*

Select an optionExecutive / C-suiteVPDirectorManagerIndividual ContributorOther

Company's AI adoption maturity*

Select an optionExploring opportunitiesRunning pilotsScaling adoptionEmbedded in the business model

Managing Without AI

What’s surprised me most about our AI-powered workflows and processes isn't AI itself; it's how I ever managed the old way!

During discovery with clients, we’re often taking in a huge volume of information: multiple departments, dozens of users, branching processes, spreadsheets, and video walkthroughs. It can feel like drinking from a fire hose.

Now, we can use secure AI project management software to organize that information, so we see both the 10,000-foot view and the fine details. It also lets us quickly reference things we’ve already discussed without having to circle back to the client.

That balance of clarity and efficiency has been game-changing. The surprising part is how quickly AI has become second nature; it's that I honestly don’t know how I worked without it.

Real-World AI Workflows That Save Delivery Time

Here's an example of AI in project management that showcases truly game-changing implementation.

One common challenge in workflows is missing information at the point of submission. Take a design team, for example, where requests can come in from any department. To do their job well, designers need details like pixel dimensions, color palettes, inspiration examples, and due dates. Without them, the process slows down.

We solved this by building an AI-powered workflow in Asana that reviews submissions before they reach a human. If key details are missing, AI flags the request, rejects it, and tags the submitter to fill in the gaps. Only once the task passes this check is it routed to a designer.

This has eliminated the back-and-forth that used to waste time. And it has protected designers from getting bogged down with incomplete requests. They now only see work that is fully prepared and ready for them, which has made the process both faster and less frustrating.

The step-by-step process of adding AI Checks

Here is a step-by-step process for using Asana's AI Studio rule builder to add in AI checks:

• Enable AI Studio
  • Make sure AI features are switched on in your Admin Console.
  • This only works on current Business/Enterprise plans (not legacy tiers).
• Create a New AI Rule
  • In your project, go to Customize → Rules → Add Rule.
  • Choose the pre-built rule "AI Studio → Check for Missing Information".
• Define the Trigger
  • We used “When a task is added to the project” (typically a form submission).
  • You can scope it to all tasks or just those from a specific intake form.
• Add the AI Instructions (the “prompt”)
  • AI Studio agents act like mini GPTs, you tell them exactly what “good” looks like.
  • Our baseline definition of “sufficient info” included:
    • Clear description of the request
    • Due date or timeframe
    • Dependencies on other work
    • We customized this further to reflect our creative process (e.g., objectives, audience, video style).
• Configure the Actions
  • If info is missing, the AI:
    • Adds a comment requesting clarity (max 3–5 questions).
    • Creates a subtask assigned back to the requestor so nothing falls through the cracks.
  • If sufficient info is found, the task simply moves forward.
• Choose the Model
  • For this use case, we used a small, low-cost model (GPT-4o Mini) since it’s just classifying and prompting, not heavy analysis.
  • Larger models (Claude Opus, Sonnet 4) are overkill unless you need cross-document reasoning or complex summaries.
• Test With a Sample Submission
  • We submitted a vague form (“Need a new video for a global campaign”).
  • The AI flagged it as incomplete, asked for campaign goals, target audience, and style, and spun up a subtask for the requester to fill in.
  • It even referenced up-to-date marketing best practices to guide the request.

That's the end of the process. It checks our submissions for missing information, and loops in the submitter if there is anything missing.

Enhancing Delivery Rituals Without Disrupting Them

AI has also impacted our delivery rituals.

Our sales team has been using AI to capture context from sales calls and turn that into a starting point for defining scope. That saves a lot of time and gives delivery a stronger foundation before the work even begins.

When it comes to keeping everything aligned, though, that’s still primarily driven by our core Asana setup. Visibility into status, priorities, and next steps is what keeps everyone on the same page, and AI doesn’t replace that.

Where AI has added real value is in summaries. Being able to generate a quick overview of a project makes it much easier to onboard new team members or reassign work when someone is out, without having to manually catch them up.

In practice, AI is becoming a support layer around our rituals rather than replacing them. It speeds up context gathering and knowledge transfer, while the human side remains critical for alignment and decision-making.

Agentic Workflows: Drafts, not decisions

Agentic outputs give us a draft structure, but real-world delivery still demands human refinement.

Share This Quote on:

• Share on Twitter
• Share on LinkedIn
• Share on Facebook

Brandon Llewellyn

Head of Delivery @ Cirface

We’ve also been doing AI experimentation with agentic workflows — in a light way.

As a first step, they can be very useful for structuring client deliveries. The challenge is that AI rarely gets it fully right when it comes to hierarchy or workflow design. And even when it does, there are always real-world situations that the AI doesn’t account for.

Because of that, we treat agentic outputs as starting points rather than finished products.

They give us a draft structure we can react to, and then we refine it with our own judgment and client context. It’s going well, in the sense that it saves us time up front, but we still see the human layer as essential for getting to a workable system.

Inside Cirface's AI-First Delivery Stack

Our core stack is built around Asana for project management, Harvest for time tracking, Slack for collaboration, Google Workspace, Fellow, and Miro. Those are the constants that keep everything moving.

Over the past year, we’ve doubled down on Fellow, especially as an AI-powered note taker. That’s been a real unlock for capturing and structuring conversations without adding overhead. In fact, if there's one tool that saves our team the most time, it's Fellow.

I’ve also started using Notebook LM in certain cases where I need to quickly digest and navigate large amounts of information.

Regarding Miro as a task management tool, its evolution hasn’t been centered on AI, but they’ve been consistently rolling out updates that make collaboration smoother and more intuitive, which has kept it valuable in our process.

Clearing the Runway: Where AI Adds The Most Value

Right now, the areas most ripe for AI support are the ones that involve gathering, shaping, and checking information. We’re getting a lot of value from AI-powered briefs, summaries, translations, overviews, and filters that ensure nothing is missing before work gets handed to a human.

These are the kinds of tasks that used to be repetitive and time-consuming, but they’re now handled in seconds with far fewer errors. Where AI is still limited is in anything that requires real judgment, creativity, or nuance.

For example, understanding the politics between departments, weighing trade-offs that affect stakeholders differently, or recognizing when a client’s ask doesn’t quite match what they actually need. That kind of reasoning still demands a human touch.

AI is best at clearing the runway: It takes care of the heavy lifting with information so that the humans can focus on interpreting, advising, and making decisions that require context and empathy.

Share This Quote on:

• Share on Twitter
• Share on LinkedIn
• Share on Facebook

Brandon LlewellynOpens new window

Head of Delivery @ Cirface

The Future of AI Delivery: Fear Versus Reality

AI will not replace project managers or delivery leads, but it will completely reshape what their roles look like. AI will take on most of the inputs, such as processing data, drafting updates, and checking for gaps, while humans will focus almost entirely on the higher-value outputs.

That means creativity, judgment, empathy, and contextual decision-making will become the core of delivery roles.

Brandon's Tip

The fear is replacement, but the reality is orchestration. AI will run the systems, and humans will remain the ones guiding, approving, and aligning work in ways that machines cannot.

Use AI Without Handing Over the Steering Wheel

My advice would be to keep testing and stay open to using AI as a way to make your work easier.

Right now, it is one of the biggest opportunities to save time and free up headspace, and I believe it will soon become a baseline requirement for most digital roles. The sooner you build familiarity, the better positioned you will be.

And at the same time, lean on your own judgment. Everyone is saying “adapt or be left behind,” but not every use case makes sense, and not every solution is trustworthy.

The goal is to use AI to enhance delivery, not to hand over the steering wheel completely.

Get in touch!

Brandon's DMs are open on LinkedIn for anyone looking to discuss further or for any advice. And check out Cirface!