Last quarter I walked into a quarterly business review with the CIO of a regional telecom carrying about $4M in committed annual spend with my employer. Two weeks of prep had compressed into one. The holiday cycle had eaten the buffer. I was sitting in the lobby trying to remember the name of their head of network operations and which of his three priorities had moved since our last touchpoint.

I opened Perplexity on my phone. Forty seconds later I had it. A fresh executive summary of his priorities, pulled from a recent industry interview. The names and roles of the three people I was about to meet. A clean read of their parent company’s last earnings call, where AI was mentioned eleven times.

The meeting went well. Not because I was clever in the room, but because I had bought back the prep time the calendar had stolen.

The five tools below are the ones that, week after week, give me back hours I do not have. Nothing exotic. Nothing that requires an engineering degree. Just the AI workflow I have actually settled into after eighteen months of trying everything.

Perplexity for the prep that has to happen in twenty minutes

Account research used to be a Google plus LinkedIn plus 10-K marathon. Now it is a Perplexity search.

The specific use case where it earns its place: pre-meeting prep. Before any new conversation with an enterprise stakeholder I have not spoken to in three months, I run two Perplexity searches. The first is “What has [Company] announced about AI or cloud transformation in the last six months?” The second is “What is [Person Name] at [Company] focused on right now?”

The output gives me linked sources I can click through if I need depth. Most of the time I need the headline and three bullet points. That is what Perplexity returns in under a minute.

What it does not do well: depth on private companies, anything that needs proprietary internal data, or finance-specific calculations. For those I am back to manual research. But for the 80% of pre-meeting prep that is “what is this person publicly saying right now,” nothing else comes close on speed.

The honest reason this beats Google: Google gives me ten links and asks me to do the synthesis. Perplexity does the synthesis and shows me the ten links if I disagree. When I have twenty minutes before a meeting, that difference matters.

ChatGPT for the email I do not want to write

I write twenty to thirty client emails a week. Most of them are minor variations on the same thing: follow-up after a discovery call, re-engagement after a stalled deal, scheduling note for an executive. The patterns are identical. The friction was that each one still required ten minutes of careful drafting because the tone matters when the recipient is a VP at a Fortune 500 telecom.

ChatGPT replaced that friction with three saved prompts. A follow-up template, a re-engagement template, and a “we missed our window, here is the path forward” template. I paste in the last few messages from the thread, ask for a draft in my tone, and edit lightly. Three minutes per email instead of ten.

The one rule I enforce: no marketing language. I tell it explicitly to write like one peer writing to another, not like a vendor writing to a buyer. The output stays usable. The recipient never knows it was drafted with help.

What I do not use ChatGPT for is anything strategic. The first conversation with a new account, the framing of a deal escalation, the post-mortem after a loss. Those need my brain, not a model’s. ChatGPT handles volume. Strategy handles itself.

NotebookLM for reading the documents nobody has time to read

When a customer sends over a 60-page security questionnaire, a regulatory filing, or a multi-vendor RFP, my old workflow was to skim and miss things. Now I drop the document into NotebookLM and ask three questions. What is this customer actually optimizing for. What are the deal-breaking constraints. Which of our capabilities map cleanly to the requirements.

The reason this works better than feeding the same document into ChatGPT is that NotebookLM only answers from the document itself. It will not extrapolate or fill in gaps from training data. For source-grounded reading where I need to cite the customer back to themselves accurately, that is the right behavior.

I also use it for earnings calls. If a customer’s CFO said something three quarters ago about AI infrastructure spend or cloud migration timelines, I want to surface that line in twenty seconds, not forty minutes of scrubbing transcripts. NotebookLM does that.

The limit is real though. NotebookLM will not catch nuance the document itself does not contain. If the security questionnaire is the polished version and the real concerns are buried in a side conversation with their CISO, no amount of synthesis will surface them. The tool is only as good as the source.

Claude for the writing that has to be right

For anything going in front of a customer in writing — an executive summary, a one-page proposal, a response to a tricky email from procurement — I draft in Claude.

The output is consistently more careful than ChatGPT’s, with better instinct for what to leave out. ChatGPT tends to expand. Claude tends to compress. When I am writing for a CIO who reads forty emails before lunch, compression matters more than completeness.

The specific workflow: I dump everything I know about the situation into a Claude conversation, give it the audience and the desired outcome, and ask for three drafts. I pick the one that opens best, then edit it down to my voice.

The reason I do not use ChatGPT for this layer is style. ChatGPT defaults to a slightly performative register that reads fine internally but sounds off when an enterprise customer reads it. Claude’s default register is closer to how a thoughtful colleague writes. For high-stakes external communication, that gap matters.

Notion AI for the account plan nobody else will ever read

Every enterprise account manager I know maintains an account plan that nobody else will read. It exists for one reason: to force me to write down what I think I know, and to reveal what I do not.

Notion AI lives inside that document. I use it for one specific job: synthesizing meeting notes into structured updates. After a customer call I dump my raw notes into the page, then ask Notion AI to extract three things — what changed about their priorities, which stakeholders moved, and what I committed to send. It produces a clean update I can paste into Salesforce in thirty seconds.

The trick is to never let it write the strategic part. Synthesis it can do well. Strategy is mine. The model has no idea why a particular stakeholder went quiet, or why a deal that should have closed last month is suddenly waiting on the General Counsel. Those answers live in conversations the model never sees.

What this stack actually changes

None of these tools are exotic. None of them require special access or a CTO sign-off. Together they save me roughly six hours a week, not by doing my job for me, but by clearing the friction that used to live between every two productive activities.

The bigger shift is psychological. Before this stack, I treated prep as the cost of every meeting. Now prep is a fixed twenty-minute slot, and the meeting itself is where I focus. The customer feels it. So does my pipeline.

One honest caveat: none of this matters if the underlying account work is weak. AI tools amplify what is already there. If your discovery is shallow, AI will help you write better emails about a shallow discovery. The tools are leverage, not substitute.

If you are an enterprise account manager wondering where to start, do not start with five tools. Start with one. Perplexity for pre-meeting prep is the highest-leverage single change you can make. Build the habit. The rest will compound from there.

Related reading

• Claude Pro vs ChatGPT Plus: A Decision Tree for Picking the Right $20
• How I Turn a Customer Sales Call Into a Case Study in Under an Hour Using Granola, Claude, and Notion
• The Hidden Cost of AI Subscriptions: My 2026 Audit (and the Framework That Saved Me $2,200/year)