How to evaluate an AI startup job offer: cash, ESOPs and runway

Step 1: compare cash, not CTC

CTC figures combine things you receive monthly with things you may never receive. Split both offers into the same components.

The startup's quoted total looks much larger. Now compute what's comparable. A short script makes the assumptions explicit:

# Hypothetical offers. Adjust the probabilities to your own judgement.
stable = {"fixed": 24_00_000, "variable_target": 3_00_000, "variable_payout": 0.8}
startup = {"fixed": 27_00_000, "joining": 1_00_000,
           "esop_paper_value_4y": 15_00_000, "esop_haircut": 0.0}

def year1_cash(o):
    cash = o["fixed"] + o.get("joining", 0)
    cash += o.get("variable_target", 0) * o.get("variable_payout", 0)
    return cash

s_cash, u_cash = year1_cash(stable), year1_cash(startup)
print(f"Stable year-1 cash:  ₹{s_cash:,.0f}")
print(f"Startup year-1 cash: ₹{u_cash:,.0f}")
print(f"Startup ESOP counted: ₹{startup['esop_paper_value_4y'] * startup['esop_haircut']:,.0f}")

assert s_cash == 26_40_000
assert u_cash == 28_00_000

(Python prints with Western comma grouping; ₹26,40,000 and ₹28,00,000 in Indian grouping.) On cash alone, the difference is ₹1,60,000 in year one, and less after year one because the joining bonus doesn't repeat. That — not the ₹16,15,000 gap in quoted CTC — is the certain premium for the extra risk. Setting the ESOP haircut to zero is deliberately conservative; change it only after Step 2.

Step 2: ask about the ESOP terms

The offer's rupee figure for options usually reflects a valuation the startup chose. Ask for the terms in writing:

If the startup cannot or will not answer these, count the ESOPs at zero.

Step 3: judge the company's staying power

The third row is specific to AI startups. A product whose only advantage is a clever prompt on someone else's model is easy to copy. Ask what the startup owns.

Step 4: weigh what you'd learn

Startups often give broader responsibility early — talking to customers, deploying to production, choosing tools. That can be worth real career value, but name it concretely:

• Will I own something end to end within six months? What?
• Who will I learn from? Have they built and shipped AI systems before?
• Will I work with customers directly?

If the answers are vague, the learning premium may be smaller than it looks.

Step 5: set your own limits

Before deciding, write down:

• The minimum cash you need to cover expenses and savings goals for a year.
• How many months you could manage if the startup failed and you needed a new job.
• Whether anyone depends on your income, and how that affects the risk you can take.

If the startup's cash offer covers your minimum and you could manage a gap, the risk is one you can choose. If it doesn't, the ESOPs shouldn't persuade you.

Negotiating

You can often improve the certain part of an offer. Reasonable asks: a higher fixed component instead of more options, a longer exercise window after leaving, or a written confirmation of the ESOP terms. Ask for everything in writing before you resign from your current role.

For experienced engineers weighing client-facing AI roles, the consultant vs product engineer temperament quiz helps with the kind of work you'd enjoy. If enterprise AI delivery is the direction, FDE for Professionals is being planned for engineers making that move; its schedule has not been announced.