Skip to content

Watch: Using AI effectively in project & delivery leadership

                                                            gonzague thumbnail


The documentation, the repetitive status updates, the consolidation of inputs from twelve different stakeholders, the drafting of communications that everyone rewrites anyway: these are the tasks that consume project management time without delivering what project management is actually for. The question is not whether generative AI can do these things. It clearly can. The question is whether you know how to direct it precisely enough to trust the output and use it to free up the work that actually requires you.

This session, led by Gonzague, a consultant and trainer with over 30 years of experience across service management, DevOps, PMO and generative AI, is a two-part practical workshop. The first part covers the current landscape and the specific pitfalls that cause professionals to use AI badly, including hallucinations, data governance failures and over-reliance on outputs they cannot verify. The second part is a hands-on session using a prompt library built specifically for project managers, applying real prompts to real project scenarios across the delivery lifecycle.

Watch the full recording here:

 

What you'll take away from the session:

  • How to structure prompts that produce reliable outputs: the Role, Task, Format framework and when to use more advanced techniques including variables for reusable prompts, chain-of-thought prompting for complex multi-step outputs and chained prompts that build on each other across a workflow

  • The most important pitfalls to understand before relying on AI in your practice: hallucinations that present incorrect information with full confidence, memory loss in long conversations as the context window fills up, training data bias, and the fundamental principle that AI does not understand what it generates, which means human review is always non-negotiable

  • The data governance rules that every independent professional and project manager must apply: why personal identifiable information must never enter a public LLM, why financial and proprietary organisational data requires a private or secured platform and how to anonymise data before processing it

  • Practical AI use cases across the full project lifecycle: drafting project charters and scope definitions in initiation, building risk matrices and WBS structures in planning, consolidating stakeholder inputs and generating status reports in execution, and extracting lessons learned and closing documentation at project close

  • A live demonstration of a project-specific prompt library, including shortcuts for project charter generation, stakeholder identification, risk analysis, WBS creation, stakeholder sentiment analysis and end-of-project documentation, applied to real anonymised project scenarios

  • How to build an AI adoption roadmap for your own practice or your clients: starting with an honest assessment of current capability, identifying champions, creating a secure experimentation environment, evaluating use cases with domain experts and integrating AI into existing workflows rather than replacing them wholesale


If you are managing projects without using AI to accelerate the parts that do not need you, this session will show you exactly where to start.

What is prompt engineering and why does it matter for project managers?

Prompt engineering is the practice of structuring your instructions to an LLM so it produces the most useful possible output. For project managers, this matters because the difference between a vague prompt and a well-structured one is the difference between a generic response and something you can actually use. The RTF framework, Role, Task and Format, is a simple starting point: tell the model who it is acting as, what you want it to do and what format you want the output in. More complex tasks benefit from step-by-step instructions that guide the model through the reasoning rather than asking it to produce everything at once. 

What are hallucinations and how do you protect against them?

Hallucinations are outputs that look correct and are presented confidently but contain information that does not exist or is factually wrong. The most reliable protection is Gonzague's rule: only use AI on topics you know enough about to verify the output. If you cannot evaluate whether what the model produced is accurate, you should not be using it unsupervised on that task. This is particularly important for regulatory or technical content where a plausible-sounding error can cause real problems. 

What data should never be entered into a public LLM?

Any personally identifiable information, confidential client data, financial forecasts, proprietary process documentation or anything that carries legal or contractual sensitivity. Public LLMs store and may expose your inputs. Gonzague's example: sales managers uploading client forecasts to free ChatGPT subscriptions is a serious data breach risk. If your work involves sensitive data, you need either a private deployment of an LLM or a process for anonymising the data before it is processed. 

How should an organisation approach adopting AI for project management?

Gonzague's roadmap has four stages. First, assess current capability honestly: what do people actually know and not know about the technology. Second, provide foundational training and a secure experimentation platform before asking people to evaluate use cases. Third, let domain experts, people who know the business problems, identify where AI genuinely helps rather than where it looks impressive. Fourth, integrate AI tools with existing systems rather than replacing workflows that are already functioning. Gonzague's core principle: adopt AI because you have a business problem, not because the technology exists. 

 
, ,