Add an AI assistant to a Delhi service desk?
Before the decision
Agents repeatedly ask supervisors to interpret scattered policy notes. Team leaders spend time correcting responses that look reasonable but omit an important policy exception. The pilot must measure what happens after a first response is sent.
The problem
Should the team choose AI-assisted agents or Searchable playbooks for the next 12 months?
Decision owner: The business unit lead.
A confident wrong answer can create costly repeat contacts.
Option economics
Scroll the table to see all figures.
| Option | Volume (resolved cases) | ₹ per unit | Fixed cost | Net contribution |
|---|---|---|---|---|
| AI-assisted agents | 5,10,000 | 16 | ₹62 lakh | ₹19.6 lakh |
| Searchable playbooks | 4,55,000 | 14 | ₹44 lakh | ₹19.7 lakh |
Make the call
- Calculate net contribution for AI-assisted agents and Searchable playbooks. Which assumption makes the comparison most fragile?
- If the leading option delivers 10% fewer resolved cases, does your recommendation change?
- What evidence would you ask for before the team makes an irreversible commitment?
Reveal the worked solution
Start by validating Searchable playbooks. It contributes ₹19.7 lakh after fixed costs and leads by ₹0.1 lakh. Treat that gap as the amount of unmodelled cost or lost contribution the option can absorb before the ranking reverses.
- Frame the decision
Compare the cost of delivering a successful outcome, including exceptions and manual work. Check whether the operating team can run the new process.
- Calculate both options
AI-assisted agents: 5,10,000 resolved cases × ₹16 less ₹62 lakh in fixed costs gives ₹19.6 lakh. Searchable playbooks: 4,55,000 resolved cases × ₹14 less ₹44 lakh in fixed costs gives ₹19.7 lakh.
- Stress-test the forecast
Searchable playbooks leads by ₹0.1 lakh in the base case. If Searchable playbooks delivers 10% fewer resolved cases (4,09,500) at the same margin and fixed cost, it contributes ₹13.33 lakh. Holding the other option constant, the stressed result falls behind. This isolates volume risk. Check whether both options would respond to the same change in demand.
- Set a decision threshold
Holding the other option’s result fixed, Searchable playbooks needs at least 4,54,286 resolved cases to match or beat it under the model. Compare audited resolution quality and seven-day repeat contacts on a staged rollout.
A confident wrong answer can create costly repeat contacts. Compare the cost of a limited test with the loss the business would face if the full commitment failed. Ask the team to agree the decision criterion before collecting more evidence.
What if the whole market moves?
The worked solution tests a change to one option at a time. Here, both options gain or lose the same percentage of volume. Unit contribution and fixed costs stay constant.
| Volume scenario | AI-assisted agents | Searchable playbooks | Higher contribution |
|---|---|---|---|
| Both lose 10% of volume | ₹11.44 lakh | ₹13.33 lakh | Searchable playbooks |
| Base assumptions | ₹19.6 lakh | ₹19.7 lakh | Searchable playbooks |
| Both gain 10% of volume | ₹27.76 lakh | ₹26.07 lakh | AI-assisted agents |
Explain which option you would take forward. Check whether capacity, price or customer mix would also change as the market moves.
Practise this case as an interview