Over the last few weeks, I’ve been tinkering with a small side project I've called "PlanSpark" - a lightweight tool that helps Customer Success Managers (CSMs) and/or their clients generate a success plan for achieving ROI on their cloud technology investments.
At its core, PlanSpark takes some key inputs like:
- The clients type of business
- The vendor's technology solution
- The subscription or modules the client has purchased from the vendor
- The clients KPI’s and goals, as pulled from a call transcript or meeting not
With that, it uses AI to generate a clean, structured JSON-formatted Success Plan — including Recommendations and S.M.A.R.T. tasks — which can be easily dropped into a project management tool or even a simple Excel file for editing and more robust tracking (i.e. adding task ownership etc.)
How I Built It
This was my first real experiment using a no-code platform (Bubble.io, in this case) with AI generation functionality, and honestly, I was impressed with how much can be accomplished quickly.
Bubble’s new AI-assisted builder feature helped jump-start the whole thing. I provided it with a prompt describing what I was trying to build, and it scaffolded out a basic front-end, database, and workflows. The result wasn’t perfect, but it was enough to start iterating right away.
From there, I focused on building the AI-powered functionality. I hooked in OpenAI’s API and designed a prompt to generate success plans based on the user’s input. The prompt analyzes the vendor’s offering, what the client currently owns, and the clients goals and creates Recommendations and Tasks to help the client achieve them.
To help manage usage, the Bubble AI added a simple credit system - where each AI-generated success plan consumes 1 credit (to help offset the cost of calling the OpenAI API). New users automatically receive 2 free credits to take it for a spin.
How to Use It
- Create your account with an email address and password and if you are a Client (organization that uses the solution) or Vendor (organization offering/selling the solution)
- Your "Dashboard" page will be empty, so go the "Clients & Vendors" page to add Clients and Vendors. All fields are required when creating a new Client or Vendor (including the addition of their logo image - I could have left this optional, but makes for a better looking dashboard when you have created a few plans)
- Go to the "Success Plan" page and enter in all the info
- Select one of the Clients you added as well as one of the Vendors
- Provide high level details about the services/modules that the Client has purchased from the Vendor
- Enter in the Client's goals for using the software, trying to be as descriptive as possible
- Select how many Recommendations and associated Tasks you want in your plan
- Use the slider to adjust the AI's level of creativity when creating the plan
- Click the "Generate Plan" button (this is when a credit will be consumed)
- Allow the API call to run (can take up to 20 seconds, you'll see the page reloading)
- When done, the "2. Plan Preview" section will be updated with your AI generated plan
- Hit the "Save & Send (JSON)" button which will save the plan to your dashboard and also email you a copy of the JSON text.
* In the event you need additional credits, you can head over to the "Credits" page.
Once you've got the JSON text for your plan, you can import it into your project/task management software (including Excel). You'll get a well structured plan with a top level Success Plan, associated Recommendations, and associated Tasks (with sample due dates). Excel does a good job of ingesting JSON files (.jsonl) and displaying them. Should you need to, you can always save this as a CSV file if needed.
Plans can be viewed, emailed, and deleted from the "Dashboard" page.
That's all there is to it. Nothing too fancy.
Why This Was Worth Doing?
To be clear — this isn’t a polished product, and I’m not gearing up to launch or promote the heck out of it and there are other, likely even easier ways to perform the same core functionality (i.e. online forms + automation + API calls). In addition, there are some quirks I'd still need to iron out:
- The AI prompt could be refined further — it works well, but some outputs could be sharper or more context-aware. Perhaps leveraging OpenAI's Deep Research API (once available) could be a starting point (?)
- The app is not fully mobile responsive — it’s best used on a desktop or tablet (in landscape)
- And it’ll still occasionally throw an API error when OpenAI gets too creative with its output (despite the constraints I've put on it). I'm sure I could put some error handling in there though.
But the point of PlanSpark wasn’t to build the next big thing. It was to learn and test drive.
I wanted to explore:
- How fast could I build a working SaaS tool using modern no-code + AI? (I had considered Replit but wanted to be able to make modifications easily and without any coding at all.)
- Could I get something usable within a few weeks?
- Could AI be embedded into the product itself, not just used to build it?
Turns out the answer is yes. And while PlanSpark is modest in size, it serves a real purpose: helping tech vendors and their clients kickstart the value planning process for their cloud tech — something I’ve always cared about (see previous posts if you don't believe me).
PlanSpark isn’t perfect — but it’s real. And that’s kind of the point.
If you’re curious about what AI and no-code can do together, I’d encourage you to try building something small. It doesn’t have to be revolutionary. It just has to work well enough to make someone’s day easier.
You'll be surprised at what you learn!
Check it out here: PlanSpark
