Most people go into salary negotiations underprepared — not because they don't care about the outcome, but because preparing properly is uncomfortable. You have to research what your work is actually worth, build a case for yourself without sounding arrogant, and rehearse saying out loud a number that feels uncomfortably large. It's the kind of preparation that's easy to defer until the night before, and then the night before arrives and you're winging it.
AI tools are genuinely useful here. Not because they can tell you what you're worth — they can't — but because they can help you research market rates, pressure-test your reasoning, draft the framing for your ask, and run through the conversation in advance. The awkward practice reps that most people skip because there's no one to practice on? You can do those with Claude or ChatGPT. They won't judge you for stumbling, and they'll push back if you ask them to.
Here's how to actually use this, step by step.
Step One: Research What the Market Actually Pays
Before you settle on a number to ask for, you need a credible picture of what your role, at your level, in your location, pays in the current market. The mistake most people make is either skipping this step entirely and anchoring to what they currently make, or pulling one number off a salary comparison site without any context.
AI tools can help you build a more complete picture, but with an important caveat: ChatGPT and Claude have training data cutoffs and don't have live access to job postings or real-time salary data. What they're good at is helping you understand which sources to use and how to interpret the numbers you find. Perplexity is worth using here too, since it searches the web in real time and can pull current data from sources like Glassdoor, Levels.fyi, the Bureau of Labor Statistics, and LinkedIn Salary.
A useful prompt to start with in Perplexity or ChatGPT:
The follow-up question matters as much as the first: ask it to explain what factors tend to push someone toward the higher end of the range — company size, specific skills, management responsibility, technical specialization. This helps you understand not just where the market sits but where you sit within it, which is the number that matters for your ask.
Step Two: Build Your Case in Writing First
Knowing what the market pays is only part of the preparation. The other part is building the argument for why you specifically deserve to be at the number you're asking for. This is where most people get vague — they know they've done good work but struggle to articulate it in terms that land in a negotiation conversation.
Try this with Claude or ChatGPT: describe your role, your responsibilities, and three or four specific contributions you've made in the past year — ideally with some numbers attached, even rough ones. Then ask the AI to help you shape this into a clear, concrete case that connects your contributions to the ask you're making.
What comes back will probably need editing — the AI's version will be a bit more formal than how you actually talk, and it won't know the specific details of your workplace or your relationship with your manager. But it will give you a structure you can make your own. The blank-page problem — sitting down to write your case and not knowing where to start — disappears.
Pay attention to whether your case focuses on what you've done or what you're worth to the organisation going forward. The strongest negotiation cases include both, but the forward-looking piece often lands harder with managers, who are thinking about budget cycles and what they're investing in. If your draft is all backward-looking, ask the AI to help you add a forward angle.
Step Three: Rehearse the Hard Moments
This is the part most people skip, and it's where AI earns its place in this process.
Salary conversations go off-script in specific ways. The manager says the budget is frozen. They say your performance has been great but this isn't the right time. They counter with a number significantly lower than you asked for. They go quiet in a way that makes you want to fill the silence with a concession. Each of these moments has a response, but if you haven't thought about them in advance, you're improvising under pressure — which is when people tend to fold.
Ask Claude or ChatGPT to roleplay the conversation with you. Set it up specifically:
Then actually do it. Type out what you'd say as if it were a real conversation. When the AI pushes back, respond. Notice where you hesitate, where you over-explain, where you apologise for the ask. These are the rough edges to smooth before the real conversation.
After a few rounds, ask the AI to give you honest feedback on how you handled the pushback. Was your response clear? Did you hold your number or talk yourself down? Did you get defensive when the manager challenged you? This kind of feedback is genuinely useful, and it's available at 10pm when you're doing your prep, not just during business hours when a mentor might be.
A Few Things AI Cannot Do for You
It's worth being clear about the limits here, because the prep work above will only take you so far.
AI doesn't know your specific workplace, your manager's personality, or the internal dynamics that shape how these conversations go at your company. It can help you prepare for a generic version of the conversation, but you'll need to adapt that preparation to the actual context. If you know your manager responds badly to formal-sounding language, the AI's draft will need softening. If you know budget conversations happen at a specific time of year and you're not in that window, that matters more than the strength of your case.
AI also can't verify salary data. Everything it tells you about market rates should be cross-checked against actual sources — Glassdoor, LinkedIn Salary, industry-specific surveys, or conversations with peers in your field. The AI is useful for helping you think about the data, not for being the source of it.
And the actual conversation is yours to have. The preparation AI helps you do is real preparation — you'll walk in better informed and having thought through the hard moments — but the outcome depends on the real conversation, the real relationship, and a dozen factors no amount of prep can fully control.
Putting It Together
The full preparation sequence looks like this: use Perplexity to research current market rates and understand where you sit in the range. Use Claude or ChatGPT to help you build and articulate your case in writing. Then use the roleplay function to run through the conversation, including the pushback scenarios, until the hard moments don't catch you off guard.
The whole thing takes two or three focused hours. That's not nothing, but it's a reasonable investment for a conversation that could move your compensation by thousands of dollars annually — and compound forward from there, since future raises and job offers often start from where you are now.
Most people who go in underprepared don't get what they're worth — not because they aren't worth it, but because they couldn't articulate it clearly in the moment. The preparation is what makes the difference. AI just makes the preparation easier to actually do.