02 · The Questions We Were Solving For
info · Part A is the Round 2 Submission brief (file
attachment.txt), verbatim. Part B is the earlier question set from the form (the one you pasted at the start), verbatim. Part C is the track and the Round 3 task, in your words.
Part A · Round 2 Submission brief (verbatim)
Round 2 Submission
Submission Deadline
Friday, 25th September, 11:59 PM IST
The submission portal is not live yet. The organisers will notify participants once it opens.
What to Submit
Answer all eight questions in the submission form.
Optionally, submit a link to the conversation or chat with the AI tool used during the process.
Round 2 Objective
Design an agent that is specified clearly enough for someone else to build.
The submission should clearly define:
- What the agent is accountable for.
- How autonomous the agent is.
- How the agent behaves across different states and scenarios.
- What existing capabilities the agent can use.
- What capabilities are missing and need to be built.
- How humans interact with the agent.
- What the agent is called.
- Which Indian company could potentially build an agent like this.
The Eight Questions
1. Agent Outcome
What is the outcome your agent is accountable for?
Answer in one sentence.
2. Level of Autonomy
What is the level of autonomy your agent has?
Score the agent from L1 to L4.
The level is determined by the most consequential action the agent takes without asking anyone.
| Level | Description |
|---|---|
| L1 | Prepares things for the person. The person still does the work. |
| L2 | Proposes a specific action every time. The person says yes or no to each action. |
| L3 | Acts independently within limits set by someone. The person handles anything outside those limits. |
| L4 | Plans several steps, executes them, checks its own work, comes back when stuck, and is judged on whether the outcome was achieved. |
3. Agent States and Flows
What are the states your agent goes through?
Describe:
Happy Flow
The normal successful journey from start to completion.
Unhappy Flow
What happens when something goes wrong, the agent gets stuck, a capability is unavailable, the user does not respond, or the desired outcome cannot be completed.
Be as specific as possible.
A diagram can be used to illustrate the flows.
4. Existing Rails and Missing Capabilities
For each of the three rails, identify:
- What your agent can leverage from the existing rail.
- What capabilities do not currently exist and must be built.
Rail 1: Gnani
Voice
Review the Gnani documentation and use the platform to understand its capabilities.
Rail 2: Pine Labs
Payments and Authorisation
Review the Pine Labs documentation.
Rail 3: Delhivery
Logistics
Review the Delhivery Developer Portal and Maps reference.
A capability table can be used to structure the answer.
| Rail | Existing Capabilities | Missing Capabilities That Need to Be Built |
|---|---|---|
| Gnani | ||
| Pine Labs | ||
| Delhivery |
5. Fourth Rail
If you had a choice, does your agent need a fourth rail to become much more effective?
If yes, explain:
- What would the fourth rail do?
- Why is it important for your agent?
- Which Indian company would you want to build it?
6. Human Interface
How will a human interact with your agent?
Describe the interface as specifically as possible.
Consider:
- How the user initiates interaction.
- How the agent communicates with the user.
- What information the user sees.
- When the agent asks for approval.
- When the agent acts independently.
- How the user handles exceptions or escalations.
- What the overall interaction looks like from beginning to end.
7. Agent Name
What is the name of your agent?
Provide the name of the agent.
8. Potential Indian Builder
Which Indian company has the best chance of creating an agent just like yours?
Name the company and explain why it has the relevant capabilities, infrastructure, distribution, data, or ecosystem to build the agent.
Judging Criteria
Submissions will be evaluated on:
- Evidence
- Creativity
- Clarity
- Feasibility
- Thoroughness
Comparative Evaluation
Submissions will be evaluated in two ways.
1. Compared Within the Same Opening
Your solution will be compared against other teams that selected the same opening.
The focus will include:
- Originality
- Insight
- Quality of the solution
2. Compared Against an AI Baseline
Each opening has also been explored using frontier AI systems.
Your solution will be compared against the organisers' internal AI generated baseline.
Important Round 2 Rules
No Coding Required
You do not need to know how to code.
The round focuses on:
- User behaviour
- Agent design
- Understanding capabilities
- Identifying gaps
- Specifying what needs to be built
Documentation Research Is Expected
Use AI to help understand the documentation of the three rails if necessary.
Missing Capabilities Are Acceptable
If your agent requires a capability that does not currently exist, that does not negatively affect the submission.
Instead, clearly:
- Identify the missing capability.
- Specify exactly what it needs to do.
- Explain how the agent behaves in the meantime.
Do Not Build Yet
Round 2 is a design and documentation round.
You are not expected to build or test the agent during this round.
The Build track will involve building in the next round.
Show Your Work
AI tools can be used.
Teams are encouraged to submit their AI logs and conversations so that human contribution can be evaluated alongside AI contribution.
Core Deliverable
The final submission should answer these eight questions clearly:
1. What outcome is the agent accountable for?
2. What is its autonomy level, L1 to L4?
3. What states does it go through, including happy and unhappy flows?
4. What can it leverage from Gnani, Pine Labs, and Delhivery, and what capabilities are missing?
5. What fourth rail would make it more effective, and which Indian company should build it?
6. How does a human interact with the agent?
7. What is the agent called?
8. Which Indian company could build an agent like this?
Part B · Earlier question set from the form (verbatim)
How your agent works, in six steps.
Six cells · one sentence each · 15 words max
What triggers it. What it already knows. What it does. Who it deals with, person or institution. What it asks its human, and when. How it knows it’s done. One sentence per step; if a step needs two, it is two steps.
04
How your solution touches each rail.
Payments · logistics · voice — one sentence each
State the role each rail plays. If a rail has no role in your opening, say so and give the reason.
05
Which rail will you innovate on?
Select one · 40 words
The rail where your solution needs something that does not exist today. Name the missing capability.
06
Which customer asset will you ask for?
30 words
The one thing the customer must hand over for your agent to work — order history, the family group, statements, contacts — and the reason they’ll agree.
07
The annexation: which use-case do you subsume next?
30 words
The use-case your solution takes over after this one, and why it falls to you rather than to whoever owns it today.
08
Which of the sixteen openings would you never hand to an assistant?
One opening · one sentence
Assume the assistant works flawlessly. Name the opening and the reason. There is no right answer.
09
Which Indian company should have built this already?
One company · 60 words
Pick a company that has the users, the data and the payment relationship your solution needs. Make your best guess why they haven’t built this yet.
Which track are you entering?
Select one · locks with your submission
Your team enters one track together. You gave a preference when you registered; this is where it becomes final. Product Strategy teams finish the competition with a fully specified build — complete flows, the segment map, the wall you hit and why. Product Build teams finish with a working prototype on a partner sandbox. Both have separate prizes and are judged by the quality of your evidence, the depth of your load-bearing rail, the specificity of your ask.
Part C · Track and task (your words, verbatim)
ok i will tell our task. product strategy track will present their solutions and test them against what each rail can actually do toda this is our track. we have chosen product strategy and not product build
note · Logistics: Delhivery · Voice interface: Gnani · Payments: Pine Labs (The Ken's Case Competition 2026, "The Great Rewiring", in partnership with Zerodha).
Judging criteria (as pasted by you, verbatim)
This is where your product strategy chops are tested. You'll have to discover a unique consumer insight around which you assemble your agent. As with most consumer products, this insight is often unexpected and surprising — sometimes it's about asking users to hand over something they own, like their payment history, or access to their WhatsApp groups. In return, you'll have to make the case for how your agent will fundamentally rewire their life and solve a deep, emotional problem they experience every day. Finally, you'll have to design your agent across the three rails any agent has to be built on if it is going to work in the real world: * Voice — how it talks to people. * Payments and authorisation — what it is allowed to do, and how it pays and gets paid. * Logistics — how it moves things in the physical world.
Every submission is read against the same five things, whichever track you enter and whether or not you write code. * Evidence. Did a real person tell you this? Interviews, recordings, and what you found that you weren't looking for while designing the solution. * Creativity. Is the solution fresh, counterintuitive, unique, and different? This is usually where AI-generated solutions fail. * Clarity. Is the solution communicated clearly? Is it specific? * Feasibility. Is it a practical solution? Can it be implemented in a reasonable way? * Thoroughness. Is the solution meticulous and detailed?
Back to 00 README Index