Inside This CEO’s Bold Push to Use AI for Cleaner Data, Faster Delivery, and Simpler PM Workflows

Marcel's Tip

I’m trying to encourage everyone to experiment, use it, and find ways to apply it to what we’re doing to get better or more efficient. A great example is that every week, we’re doing a show-and-tell demo and celebrating something that is being experimented with around AI internally.

I'm trying to encourage everyone to experiment, use it, and find ways to apply it to what we're doing to get better or more efficient.

A great example is that every week, we're doing a show-and-tell demo and celebrating something that is being experimented with around AI internally.

Efficiencies in Context Gathering and Delivery

Context gathering has probably been the most significant efficiency we've gained with AI so far.

We've worked on centralizing context from our clients — calls, notes, Asana updates, emails, etc. — to make it all queryable with AI. Now, it's so much easier to get people up to speed on what's happening, prep for meetings, and make decisions with the right context in mind.

This saves a lot of time on internal comms, briefing, watching call replays, and drafting updates.

We're also able to save a lot of time on delivery by using AI to query the technical documentation for our systems and use it to help us write data cleanup rules and other technical tasks.

By using AI to serve as our "perfect memory", we're able to focus on and improve the customer experience. Every client now feels like our only client.

Removing Complexity from PM Workflows

Another big shift for us has been moving entirely away from bottom-up resourcing and from highly-detailed, task-based project management. The ability to create documentation that can be interacted with in a conversational way means that we can really simplify the templates and remove a lot of the detail and complexity from our project management workflows, which dramatically reduces the amount of time and effort required to spin up projects and maintain our PM tools over time.

It also significantly reduces drag when changes to our processes happen.

I think this could mean moving PM into Notion at some point in the future. But for now, we are quite happy with Asana.

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.

Phone

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

How are Clients Using AI for Creative Delivery?

As far as our clients, we're seeing AI in all the places you'd expect.

• Gathering context from the initial intake process and sales process to inform strategy and briefing.
• The ideation/brainstorming process and drafting initial concepts and briefs for the internal team.
• Rapidly mocking up concepts, especially in the initial pitching stage (this applies across copy, UI/UX, graphic design, video, image generation).
• Image generation
• On the digital marketing side of things, there is a lot of AI being applied to data and campaign analysis and helping monitor accounts/suggest edits.
• Creating variations on creative and copy.
• Creating a lot of leverage when it comes to things like SEO optimization — especially things that are more on the basic hygiene side of things, like meta tags/descriptions, etc.

The Misunderstood Importance of Data Hygiene with AI

I was really surprised by how important data hygiene and structure is when adopting AI into delivery workflows.

I think there is a fallacy out there that AI will lessen the need for investments in data infrastructure, but from our experience, it’s only making those investments more important.

Share This Quote on:

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

Marcel Petitpas

Founder and CEO of Parakeeto

I think there is a fallacy out there that AI will lessen the need for investments in data infrastructure, but from our experience, it's only making those investments more important.

Most companies will be limited in their ability to leverage AI by the quality of their underlying data systems.

Here are some things we use AI to clean:

• Correcting inconsistent employee or project names (e.g., nicknames, misspellings, prefixes/suffixes like [ARCHIVED]).
• Stripping out unnecessary noise such as extra whitespace, header/footer rows, or unused columns.
• Standardizing formats for dates, numbers, and text to align with internal reporting structures in our digital asset management software.
• Mapping time entries to the correct Payroll Grid role categories or Project Grid projects, even when the raw names don’t match exactly.
• Handling retainer scope changes by splitting projects into multiple versions based on start and end dates.
• Reassigning older or incomplete data to placeholder “historical” projects to preserve reporting accuracy without clutter.
• Extracting codes, tags, or IDs embedded in text fields using regex or pattern matching.
• Applying math adjustments to logged time (e.g., padding contractor hours, reallocating fractions of time).
• Renaming or remapping columns so exports from different tools fit a consistent schema.
• Combining or splitting fields (e.g., merging first/last name columns into a single employee name).

And that's just to name a few.

By creating really great documentation on how to write and configure these data cleanup rules, we've enabled non-technical employees to do far more of the groundwork on their own, in some cases implementing these pipelines without needing to pull in any technical resources at all.

Detailed Documentation with Notion AI

The crux of this was writing detailed documentation to inform an AI model.

Then, we have to keep it up to date. But since the documentation lives in Notion, and we're using Notion AI to query it, it's sort of a self-sustaining document. And when something new needs to be added, we can often have the AI draft the updates and paste them right in.

And the great thing about Notion is that it's pretty easy to get context from there to places like Gemini, Notebook LM, or ChatGPT. ChatGPT's agent functionality also works with Notion.

How to Enable Non-Technical Employees with AI

I mentioned enabling non-technical employees with AI — that has been the biggest unlock of all. As long as someone can express what they want to achieve in a logic statement, they can essentially write Python/SQL and build functioning data pipelines.

As long as someone can express what they want to achieve in a logic statement, they can essentially write Python/SQL and build functioning data pipelines.

Share This Quote on:

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

Marcel Petitpas

Founder and CEO of Parakeeto

Of course, there are limitations, and those limitations tend to be at the start and end of these technical tasks.

For example, making good decisions about what technologies to use, how to architect things, and what kind of edge cases and requirements need to be thought of is typically a place where a technical resource can add a lot of value — even if that means reviewing and adding some additional detail to the thinking an AI model has already done.

From there, the thing that is very difficult for a non-technical user to accomplish is setting up the infrastructure and initial workflows for a net-new process and taking something from idea to launch. In some rare cases, they can do it from end to end, but having a technical user set up the foundation is generally still a requirement, or at least something that will be far more efficient and lead to a better system and outcome.

Another spot where there is still technical intervention required at times is around troubleshooting edge cases. But as we document the solutions for this, AI assistance is enabling non-technical users to get unblocked more and more.

And once the rails and process are down, non-technical users can accomplish way more than they could a year ago.

AI Isn't Just Changing Processes; It's Changing how Processes are Created

In some ways, I think AI is helping us develop a process around things without having to develop a process in the traditional sense.

Prior to having AI agents and tools to help facilitate getting these things done, someone had to sit down and create a very detailed and often very deterministic set of steps to complete a task.

The challenges with that are:

• It's time-consuming
• It requires someone to have an accurate understanding of what steps to take
• It doesn't typically work well for very complex, nuanced, or fluid things
• It's often hard or expensive to do when a process is new, or not that mature, and likely to change
• It can slow the pace of progress by creating operational drag

With AI, someone can just create a new chat, agent, or custom GPT and start working while building process, knowledge, and context as they go. It's similar to working with a human, but it has perfect memory, and can turn around and share what they learned with the next person, instantly.

I think this is one of the most powerful things about AI — in some ways it's completely changing the way processes get created and maintained.

My AI Stack for Project Delivery

Our core stack is:

• Asana - As our Project Management Software
• Harvest - Time Tracking Software
• QBO - As our Accounting Software
• Parakeeto - For Management accounting, Forecasting, Utilization, ABR / Project performance

And we also use:

• Google Suite (Drive, Sheets, Docs, Calendar, Mail) - Cloud Collaboration Software
• Notion - Document Collaboration Software
• Miro - Flowchart Software
• Loom - Asynchronous Collaboration Software
• Fathom - Note Taking Software
• Zapier - Process Automation Software

Two Tools Every PM Needs

Notion and Zapier, in particular, have been game changers for us.

I touched on this earlier, but Notion is a platform we've really started to double down on. It's becoming a place where we can push all kinds of context and information about our clients in an automatic way, and then query it or use it as knowledge in other AI models.

It's also starting to roll out lots of tools that are making this much easier (Mail, Calendar, Meeting notes), and we may very well end up moving more and more functions into Notion over time — eventually, maybe even our project management and CRM.

Zapier has really done an excellent job at integrating AI agents into its tool. While everyone is focused on Gumloop and N8N, I would argue that Zapier is better than both of those, given the breadth of their integrations.

We have a few automations — mostly on the S&M side — that are saving us a lot of time each month in terms of automating marketing and making sure data is moving to the right places, with the right level of enrichment.

A Real-Life Zapier Automation

One of the meatier Zapier automations we have going on right now involves taking new leads that download our toolkit and doing a number of things with them:

1. Assessing them to see if they are using a work or personal email.
2. Filtering for those using a work email.
3. Extracting the company name from the email and splitting their first and last name apart.
4. Using an enrichment service to look up basic information about the company and person.
5. Applying more filters to determine if the lead is "qualified" based on a set of criteria — in a specific industry, certain number of employees, etc.
6. Doing more enrichment on the leads that are a good fit.
7. Summarizing and cleaning up some info about them, like company descriptions.
8. Adding those leads to our CRM and applying the appropriate tags/filtering.
9. Adding those leads to other automations that do things like enroll them in the lead magnet they downloaded and add them to an email drip campaign associated with that lead magnet.

A Simpler Future for PMs

I think (and hope) that project management will become far simpler in five years.

In the same way that an AI agent can enable an organization to develop and iterate on a "process" without actually having a "process," I hope that AI will enable project managers to do their thing without having to constantly be working through hundreds of detailed objects in a project management tool.

It feels like so much of a PM's time is spent rearranging tasks, timelines, time estimates, subtasks, milestones, dependencies, assignments, etc., etc., etc.

In some ways, these objects are like the programming language of the PM, and they use it to try and code a representation of what's happening at a given moment in time. In some cases, to help them make sense of things, and in many other cases, so that other stakeholders can stay informed as well.

That system feels incredibly inefficient and high-friction. Like running in sand.

I think AI will start to reduce the amount of precision and detail that PMs need to interact with to communicate and manage their projects.

Focus on the AI Fundamentals

It might feel like things are moving fast — and in some ways they are — but remember that we're a couple of years into a technology disruption that will be playing out for at least another decade or two.

Focus on the fundamentals — those won't go out of style. And look for ways to use AI to help you get better and faster.

Follow Along

You can follow Marcel on LinkedIn. 👉 Subscribe to the Digital Project Manager and keep up with everything shaping AI in delivery.