AI Automation for Project Managers: Tools I Actually Use

Let me get one thing out of the way first.

AI doesn’t manage projects. It doesn’t build stakeholder trust, read the room in a governance meeting, or make the call when two senior sponsors want incompatible things. It doesn’t know when to push back, when to go quiet, or when to escalate something that nobody else wants to escalate. That’s still your job.

What AI does — when you use it properly — is remove a significant chunk of the low-value, time-intensive work that sits around the edges of project delivery. The documentation. The drafting. The synthesis of information you already have. The mechanical parts of keeping a program’s comms infrastructure running.

I’m Google AI Essentials certified and I’ve been integrating AI tools into my delivery work across PM, ops, content, and CRM pipelines. In this piece, I’ll walk through exactly what I use, what it does in practice, and — just as importantly — where it stops being useful.

This isn’t a listicle of 20 tools. It’s the actual short list of what I reach for.

The Categories of PM Work Where AI Earns Its Place

Before I get into specific tools, it helps to name the categories of work where AI genuinely pulls weight versus where it’s more noise than signal.

The useful categories, in my experience, are:

Documentation and reporting. Writing first drafts of things: status updates, meeting summaries, project charters, RAID logs, lessons-learned writeups. These are time-consuming to produce, formulaic in structure, and often low-differentiation in content. AI is well-suited to producing a strong first draft from your notes.

Communication drafting. Stakeholder emails, briefing documents, change communication, executive summaries. The structural thinking is yours — AI handles the shaping and wordsmithing.

Research and synthesis. Pulling together context on a new project domain, summarising vendor options, identifying risk categories for an unfamiliar delivery environment, reviewing a contract or document for issues. AI is fast and broad here; it’s not a substitute for expert judgment, but it’s a useful starting point.

Workflow automation. Task creation, status synchronisation between tools, notification routing, alert logic. This is where the more technical automation platforms come in — not LLMs, but workflow tools like n8n and Zapier.

These four categories cover a large portion of the peripheral work that surrounds delivery. The delivery work itself — governance, decision-making, stakeholder management, escalation, risk judgment — AI doesn’t touch.

The Tools

Claude and ChatGPT (GPT-4)

I use both. Claude for longer-form drafting and analysis work — it handles extended context well and its outputs are cleaner for professional communication. GPT-4 for quick queries, brainstorming, and tasks where I want a slightly different angle.

In practice, I treat these as a capable first-draft colleague who writes fast, never complains, and needs to be reviewed before anything goes out. That framing matters — the review step is not optional.

What I use them for:

Status updates and reports. My weekly practice is to keep a running note throughout the week — bullet points of what happened, decisions made, blockers, next-week priorities. On Friday, I paste those notes into Claude with a prompt like: “Draft a weekly status update from these notes. Audience is the program sponsor. Tone: direct, not padded. Format: short summary paragraph, then a brief section each for progress, decisions, blockers, and next week.” I review and edit the output before sending. This saves me roughly 45 minutes a week on a typical program.

Meeting notes. After a governance meeting or working session, I’ll paste my rough notes (or a Loom transcript — more on that below) and ask for a structured summary: decisions made, actions assigned, open items. The AI doesn’t attend the meeting; it shapes the notes I’ve already taken.

Stakeholder communications. When a project hits a significant change — scope creep, a delayed delivery date, a resource conflict — I need to communicate clearly to people who are busy and not always sympathetic. I’ll draft the key points myself, then ask Claude to help me structure and tighten the message. I always rewrite from the output, not copy-paste from it.

Project charters and briefing documents. On a new engagement, I’ll use AI to produce a first-pass charter or brief from the discovery notes. It’s a scaffold, not a finished document — but getting from blank page to structured draft in 30 minutes rather than two hours is a real efficiency.

I want to be clear about one thing: I write the brief, I make the decisions, I own the document. AI produces the first draft faster. That’s the entire value proposition. If you are sending AI output to a client or sponsor without reviewing and editing it, you are not using it properly — you’re outsourcing your judgment.

Risk identification. On a project where I’m entering a domain I haven’t worked in before, I’ll ask ChatGPT or Claude to enumerate the common risk categories for that type of project. Not as a definitive risk register — as a prompt list to check against my own assessment. It catches things I might not have thought of.

n8n and Zapier

These are the automation layer — not AI in the LLM sense, but workflow tools that connect your project tools and remove the manual steps of moving information between them.

I use n8n for more complex, customised automation (it runs on your own infrastructure if needed, which matters in some client environments). Zapier for simpler, faster connections between SaaS tools.

What I use them for:

Task creation from communication. A standard pattern: when a specific type of email arrives (flagged by a rule, or by a label I apply), a Zap fires that creates a task in the project tool with the key details pulled from the email. This removes the manual “read email, open project tool, create task, copy details” loop.

Status synchronisation. When a task changes status in one tool, the workflow updates the corresponding record in another. Not every project environment needs this — but on programs where the client is working in one system and my team is working in another, automated sync prevents the “which tool is the source of truth?” problem.

Notification routing. Rather than everyone on a program receiving every notification from every tool, n8n workflows route specific alerts to the right person via Slack or email. Governance notifications to the sponsor. Technical blockers to the tech lead. Weekly summary digests assembled and sent automatically.

Document generation triggers. On some projects, I’ve set up automations that fire when a project milestone is reached — automatically generating a status email template, or creating the next sprint’s folder structure, or populating a template with the relevant project data. These save the mechanical setup time at the start of each cycle.

n8n has a steeper setup curve than Zapier, but it pays back quickly on anything that repeats. If you’re running a program with 12-week cycles and the same administrative steps fire at the start and end of each cycle, automation is straightforward to justify.

Loom

Loom isn’t strictly an AI tool — it’s an async video platform — but it’s AI-enhanced in ways that matter for project communication. Every Loom I record is automatically transcribed, has AI-generated chapter markers, and produces a searchable text summary.

Why this matters for PM work:

Written status updates don’t always land. Long documents don’t get read. But a three-minute Loom walkthrough of the programme dashboard, with a voiceover explaining what the numbers mean and what I’m doing about the one red item — that gets watched.

I use Loom for:

• Weekly async updates to sponsors who are time-poor
• Walking through a complex change before a governance meeting, so the meeting can focus on the decision rather than the explanation
• Onboarding new team members onto a project’s context and history
• Capturing decisions and rationale in a way that’s more nuanced than a document

The AI transcription means every Loom is also searchable text. I can paste the transcript into Claude for a written summary, or keep it as the meeting record. It removes the choice between “video update” and “written record” — you get both.

Google AI Essentials — What the Certification Adds

I completed the Google AI Essentials certification as a way to build a structured foundation under the practical use I was already doing. The course covers machine learning concepts, responsible AI, using AI in workplace workflows, and evaluating AI outputs critically.

What it actually changed for me was the discipline around critical evaluation. It’s easy to trust AI output when it reads confidently. The course makes explicit that confident-sounding outputs can still be wrong, incomplete, or poorly calibrated to context — and that the human review step is not a formality, it’s the whole point of keeping a human in the loop.

In a PM context, that matters particularly for risk identification and assumption documentation. If I’ve asked an AI to enumerate risks and it hasn’t flagged something significant, I need to have caught that. The certification reinforced the habit of treating AI output as a starting point, not an answer.

How AI Fits Into Each Delivery Phase

I run delivery across four phases: Frame, Map, Drive, Hand Off. Here’s where AI slots in across each:

Frame (scope, stakeholders, initial planning)

• Draft the project charter from discovery notes
• Research the delivery domain for risk categories and common constraints
• Produce first-pass stakeholder maps and comms plan templates
• Summarise input from multiple stakeholders into a coherent brief

Map (planning, resourcing, governance setup)

• Draft governance documentation (terms of reference, RACI frameworks)
• Build out the communications calendar template
• Identify dependencies and risks against the delivery plan using AI-assisted review
• Set up workflow automations for the ongoing reporting cycle

Drive (delivery, cadence, stakeholder management)

• Weekly status update drafting (30–45 minutes saved per cycle)
• Meeting note structuring
• Change communication drafting when the plan adjusts
• Automated task sync and notification routing via n8n/Zapier

Hand Off (transition, documentation, close-out)

• Lessons-learned synthesis from project notes
• Final report first drafts
• Handover documentation from accumulated project artefacts
• Retrospective facilitation prompts

The Drive phase is where AI saves the most time over the life of a project — because it’s the longest phase, and the documentation and communication work repeats every week.

What AI Cannot Replace

This is the part that matters most for credibility, so I’ll be direct.

Stakeholder trust. Trust is built through the quality of your judgment calls, your track record on commitments, and your ability to be honest under pressure. No AI tool contributes to this. The PM who automates their status updates and spends the saved time on better stakeholder conversations is in a stronger position than the one who just gets reports out faster.

Political navigation. Every complex program has a political environment — competing priorities, historical tensions, personality dynamics, unspoken agendas. Reading that environment accurately and moving through it effectively is a deeply human skill. AI has no view of it.

Real-time judgment calls. When a crisis hits — a key resource resigns, a critical dependency fails, a sponsor loses confidence — the response requires human judgment, experience, and communication ability. AI cannot make those calls and cannot substitute for the years of experience that inform them.

Accountability. The PM is accountable for delivery. That accountability doesn’t change because some of the documentation was drafted by an AI. If anything, it concentrates accountability — because the PM has removed excuses about administrative burden and has to own the output.

Nuanced communication in difficult situations. When a project is in trouble and you need to communicate clearly and credibly with a worried sponsor, a template won’t do it. The conversation requires presence, honesty, and the credibility you’ve built over time. AI can help you prepare; it cannot have the conversation.

The PMs who get the most from AI tools understand this clearly. They use AI to clear the administrative overhead, and they invest the recovered time in the parts of the job that require human judgment — which is where project outcomes are actually determined.

How to Start If You Haven’t Used These Tools in PM Workflows

If this is all new, the entry point is simpler than it looks.

Start with one document. Take the next status update you need to write. Write your bullet points as you normally would. Then paste them into Claude or ChatGPT with a clear prompt: who the audience is, what format you want, what tone. Review the output. Edit it. Send what you’re happy with. That’s the whole practice.

Then automate one recurring task. Identify one thing you do manually every week that follows a consistent pattern. A Zap or n8n workflow that takes 30 minutes to set up and saves you 15 minutes every week has paid itself back in four weeks. Start there.

Build the review habit first. Before you use any AI-generated content professionally, establish the discipline of reviewing it. Read every output. Ask whether it’s accurate. Ask whether it sounds like you. Ask whether it would serve the recipient well. That discipline protects you and everyone you work with.

The tools are genuinely useful once you’ve built those habits. The risk is treating them as output machines rather than draft machines — and that risk is entirely managed by keeping the human in the loop.

The Actual Value Proposition

There’s a version of the “AI in PM” conversation that’s all about transformation and disruption. I’m not interested in that conversation.

The actual value proposition is this: a program manager who uses AI tools well can handle a higher workload without proportionally longer hours. The documentation stays current, the communications go out on time, the synthesis work happens faster. The time that was going into administrative production goes back into delivery.

That matters for contract work specifically. When I’m engaged on a complex program, the client is paying for delivery outcomes — not for the hours I spend shaping a status update. Tools that compress the administrative work without compromising its quality are directly relevant to the value I deliver.

The tools I’ve described here are mature enough to use in professional delivery right now. They’re not perfect, they require judgment to use properly, and they don’t replace the skills that make a senior PM effective. But they’re real, they’re in my stack, and they earn their place.

If you’re building a program where AI automation in PM workflows is part of the picture — whether that’s automating reporting infrastructure, standing up CRM and ops pipelines, or just thinking through where these tools fit — I’m available for contract engagements across Australia, New Zealand, the USA, and Mexico.

Get in touch →

---

Aaron Darke is a Senior Project & Program Manager with 25+ years’ experience running complex programs across digital transformation, post-merger integration, and brand strategy. He is Google AI Essentials certified and lists AI & Automation as a core specialism. Available for contract engagements.