When it comes to AI, it’s easy to get distracted by all the flashy demos and headlines that are constantly splashed across social media. But the fact is, most people still aren’t leveraging AI in one of the simplest, most impactful ways it can be used: by creating an AI project assistant.

An AI project assistant can help you manage a wide range of projects across a wide range of business environments. It can dramatically boost your productivity and efficiency. Best of all, it’s much easier to set up than you might think.

In this article, I’ll walk through our simple, three-step approach to building an AI-powered project assistant, giving you the tools you need to create an AI assistant that will set you and your business up for success.

What is an AI project assistant?

Broadly speaking, an AI project assistant uses a generative AI platform like ChatGPT, Claude, or Gemini to manage a specific project within your organization.

For example, in our roles at HubSpot, some of our core projects include overhauling how we create demand, driving internal transformation via AI, and augmenting our go-to-market strategy. Each of these projects is associated with a desired outcome, and each one comes with lots and lots of people, communications, and all sorts of other data.

In our case, we’re managing huge teams with hundreds of people, but projects don’t have to be that massive to benefit from an AI project assistant. Whether you’re leading a thousand-person department or running a small business with just one or two employees, an AI-powered project assistant can help you act on the extensive data that is associated with your projects.

And how does it do that? There are three core parts of an AI project assistant: your context, your templates, and your instructions. Below, we’ll go through each of these vital components, sharing best practices and recommendations to help you set up an AI assistant like a pro.

1. Upload context data.

The foundation of your AI project assistant is all the structured and unstructured data you have access to that’s associated with your project. As such, the first step to build out your assistant will be to upload all of that context.

That means all your Google Docs, your Google Slides, reports, strategy documents, meeting transcripts … Any time someone on your team creates a document that’s relevant to your project, it should go straight into the AI’s context.

Importantly, this isn’t just your own data. You can collect meeting transcripts from meetings that you yourself couldn’t attend, so your AI assistant knows about every conversation related to the project. You can also include transcripts of Loom videos, as well as any other resources your organization has created or accessed, from internal reports and in-depth external research to important emails or Slack messages.

Ultimately, the context data is your chance to leverage all of the incredibly valuable data that your organization has created related to your project. So, when in doubt, upload! The AI assistant will use all those files to get smarter about your business, empowering it to offer recommendations and support that are tailor-made for your unique situation.

2. Build your templates.

The next step is to build the templates that the AI assistant will use to format its responses. This is where you define how you want the AI to present information to you for the common asks that you’ll have.

There are a few standard templates that we’ve found can be especially helpful. One is a Weekly Blockers template. Weekly Blockers is a list of all the issues or obstacles that are currently slowing down a project, and that should be solved within the next week. We use this template to get a summary at the start of each week, so we know what we need to do to continue to build momentum that week.

Another useful template is a Monthly Status Update. This template is designed to drive accountability: It lays out what we said we’d do over the last month, where we fell short, what the reasons were for those missteps, and what the plan is going forward.

Other common templates include Executive Memos (a summary of key metrics, deliverables, and status updates for a project), as well as Biweekly Momentum Drivers (docs that share what we shipped in the last two weeks, any items we said we’d ship but didn’t, and what we plan to ship in the next two weeks).

Once you’ve developed these key templates, you can upload them into the project files alongside your context data. Then, whenever you ask the AI assistant to complete one of those standard tasks, it will respond using the exact format and structure you’ve specified in the template.

3. Define your instructions.

Finally, once you’ve uploaded your context data and templates for frequent tasks, it’s time to add some instructions. The instructions are where you specify how the AI project assistant can best help you with your project.

For example, you can start by asking it to be clear, concise, and insightful in its responses. You can ask it to prioritize clarity. However, don’t sacrifice details or accuracy when nuance matters. You can ask it to surface blind spots. You can ask it to always back up its recommendations with evidence. You can even ask it to cite specific documents from your context data when making a recommendation.

This last instruction is especially important because it can help ensure that the assistant doesn’t hallucinate, or make up things that aren’t true. By telling the AI assistant to explain its recommendations and pull specific data from your documents, you can verify that its suggestions are informed by your situation.

In addition, beyond just responding to your prompts, you can also ask the AI project assistant to be proactive in identifying risks, missed opportunities, or potential second-order effects. You can instruct it to challenge your thinking respectfully, to point out when your assumptions or logic may be flawed, and to offer (and explain) a better alternative.

In other words, you can encourage the assistant to act as a real strategic partner. You can ask it to bring in external perspectives where they are valuable, and to connect patterns and trends it finds in your internal information with external insights. You can also ask it to show you overlap across teams, or places where disconnected teams may be working on the same thing without realizing it.

Part of what makes AI so powerful is that it has the ability to spot patterns that are really hard for humans to see. By explicitly instructing it to find those trends — and to turn them into actionable recommendations with clear steps and options for you to consider — you can build a tool that will uncover valuable hidden insights and work with you exactly how you want.

The Future of AI Project Assistants

With the three steps we’ve outlined above, you’ll be on your way to creating an incredibly impactful AI project assistant. But what’s next in the world of AI-powered productivity tools?

There are a few additional features that we expect AI providers like OpenAI, Google, and Anthropic may consider developing in the near future. For example, the ability to capture all emails and Slack messages and automatically include them in the context data would be really helpful.

Similarly, if these tools were able to ingest Google Docs dynamically, as they are updated in real time (rather than needing us to remember to re-upload documents whenever we change them), that would also make the process even more efficient.

But really, all of that would just be a bonus. Already, the AI tools available today make it possible to increase your productivity and work more efficiently than ever before. As long as you upload all of the relevant context data, build out the right templates, and define clear instructions, you’ll be on your way to unlocking incredible new value and supercharging your productivity.

To learn more about lead-scoring tactics and marketing growth strategies, check out the full episode of Marketing Against the Grain below:

This blog series is in partnership with Marketing Against the Grain, the video podcast. It digs deeper into ideas shared by marketing leaders Kipp Bodnar (HubSpot’s CMO) and Kieran Flanagan (SVP, Marketing at HubSpot) as they unpack growth strategies and learn from standout founders and peers.