#42 - The 2025 AI Roundup
Prior to this year I used ChatGPT sparingly. Maybe twice a month to help me polish emails? But never really integrating it into my daily workflows.
In 2025, however, the doomer narrative around AI leading to mass unemployment became compelling. As career insurance, I went “all in” and bought the $20/month GPT plan to see how much efficiency there is to gain in the GTM space.
The goal was simple: become fluent enough with AI to understand where a sales guy can still create value in a future where machines make perfect decisions, execute instantly, operate around the clock, and never need healthcare.
The good news is after a full year of using it every single day across sales, GTM strategy, RFPs, writing, analysis, content creation, and personal tasks, the reality is far more straightforward:
AI massively enhances your existing skill set, but it does not replace the underlying expertise or judgment required to do the work well.
It accelerates productivity, it improves output quality and it eliminates a huge amount of drudge work. What it does not do is operate autonomously.
Yes, these tools will enable leaner and more productive teams, but there will always be a human in the loop. It behaves like a useful idiot: extremely capable when guided well and almost useless when not.
The year-long experiment showed how powerful this technology is once you know where it truly adds value. When it cuts time AND raises quality it feels like cheating. It changes how you operate.
That realization pushed me to expand beyond GPT and become fluent across other models too. And now, heading into 2026, I want to crystallize the workflows that delivered real value and start cutting the ones that did not.
Why I Use Multiple Models
I see some of these companies touting benchmarks for how good the model is, but haven’t found them to be terribly accurate. Gemini 3.0 is technically the “best” model for some reason, but I continue to find it the least useful LLM between the 3 majors. I experimented with Perplexity and Grok, too, but never really found them to add anything I was missing. Point being - just experiment until you find what works…
Brief breakdown:
ChatGPT: best writer, best tone, best thinking partner
Claude: best integrated knowledge worker because it can access my Google Workspace
Gemini: Is helpful in Google Sheets when I need to implement complex formulas, but is weakest at writing. Don’t understand the hype…
Lovable (Honorable Mention): best example of “AI enhances expertise, but cannot replace fundamentals”
No model does everything well, so the real advantage comes from knowing which to use and when.
ChatGPT: The Best Writer, The Most Human, The Most Reliable Partner
If I had to keep only one tool, it would be ChatGPT. Nothing else comes close on tone, clarity, or the ability to turn raw thought into polished output.
Where ChatGPT excels
Email polish & Exec Communication - the best by far
High-quality RFP and proposal creation - don’t try it in one shot though, make sure you provide plenty of context and build very polished answers question by question
Marketing Assets - Can create Newsletter Articles, Case Studies and Executive Summaries with pretty much one prompt with enough context
Slide copy (not slide design)
Long-form thinking and ideation - helps me work through deal dynamics, best next actions for follow up, etc. if you assign it to be your CRO
And one of its biggest advantages is something people don’t talk about enough:
Voice input: the most underrated unlock of ChatGPT
Voice input lets you get every thought out in real time without the self-editing that naturally happens when you type. It is faster, more honest, and far more complete.
This is especially valuable for thorny emails, strategic thinking, or anything that requires nuance. Rather than staring at a blank screen trying to find the perfect opening line, I talk through the situation out loud. All the context, all the caveats, all the emotion. ChatGPT then turns that into a clear and professional message that gets the tone right.
The same applies to longer writing. The backbone of this article came from walking around with my phone, talking through ideas as they came. When something is fresh in your mind, you can capture it immediately instead of hoping you remember it later.
Voice is also a massive differentiator for GPT over Claude. Claude can handle structured prompts extremely well, but it does not have the same native ability to take jumbled thoughts and build a coherent and well structured output. GPT bridges the messy gap between raw thinking and polished communication in a way that other models currently cannot.
Typing feels efficient until you try voice. Then you realize how much friction you have been tolerating. When AI hardware matures, I believe voice will be the central experience because it unlocks the true speed and fidelity of thought.
Where ChatGPT still struggles
It cannot design beautiful slides (copy is great, formatting is not)
Image generation is fine for charts, not great for anything else
It is siloed from my work systems and the copy/pasting process is becoming more inefficient, especially as I increase my Claude usage
Net: ChatGPT remains the single best model for thinking, writing, and communicating. If it ever gets deep native integrations into enterprise workflows the way Claude is currently winning, it will dominate everything.
Claude: Slightly Less Natural Than ChatGPT, But A Far Better Operator
Claude’s writing quality is excellent. It is slightly more structured than ChatGPT, but very strong. However, its real value comes from one thing:
MCPs give it access to my entire Google Workspace.
I’ve tuned it so it only focuses on work tasks, and it behaves like a true execution engine. GPT has released this recently, but its not as advanced.
What this enables
1. Sales call follow-up
Every transcript gets emailed to me. I tell Claude:
“Go into the transcript from today’s call and draft a comprehensive follow-up.”
It pulls the email, drafts the follow-up, and gets me 95% of the way there in 10 seconds.
2. Proposal creation
“Based on the scope we discussed in this transcript, and referencing the proposals we’ve built, draft a full project plan and SOW.”
It pulls everything needed and produces a coherent structure.
3. Re-engaging old opportunities
“Review all communication with this person and draft a tailored end-of-year reconnection email.”
This increased my output 2–3x without sacrificing personalization.
I want to truly emphasize how transformational this is. Before Claude, I took notes by hand on every sales call, constantly worried I would miss something important. That distracted me from listening deeply, asking better questions, and driving the conversation forward. Then I would spend 30 to 60 minutes after each call reviewing those notes and manually crafting a follow-up email that was never as strong as it could have been.
Scoping calls were even more painful. I was juggling both note-taking and trying to synthesize complex scope details in real time. Producing a project plan or proposal afterward often consumed an entire day of focused work.
Now that friction is gone. Claude captures the conversation, pulls relevant context from past work, and produces better output than I ever could manually. A follow-up email that used to take an hour is done in less than a minute. A proposal that used to take a full day gets drafted to 80 percent completeness in one or two hours.
This represents a 3 to 5 times increase in speed with a meaningful improvement in quality. It has completely reshaped how I spend my time and what I am able to accomplish in a given week.
Strengths
Exceptional at context-specific work
Best and most flexible integrated workflow of the three
Great professional tone
Pulls from real assets, not just a chat window
Weaknesses
Slightly more formal than ChatGPT
Still cannot design slides
Still needs refinement passes to sound more human
Net: ChatGPT is the better writer, but Claude is the better worker. It is the only model that can actually shorten multi-hour workflows because it can retrieve and combine work assets on command. Honestly, I’m likely to double down here as I expect it to only get better on it’s current shortcomings.
Gemini: Overhyped
I wanted Gemini to be great and supposedly Gemini 3.0 is best-in-class on benchmarks. We have an enterprise license at work and it would save me a bunch of dough if it was better. But in practice:
Writing quality is brutal and continues to be brutal…
Emails sound robotic
Slide copy is mediocre
RFP sections feel flat and lack nuance
Refining drafts inside Gmail makes things worse, and has horrible UX issues
I often paste Gemini drafts into ChatGPT or Claude and immediately get better results
Where Gemini is actually strong
1. Google Sheets
If I need a complex formula or multi-step logic, Gemini is extremely good at generating and applying it directly if I’m building a complex ROI model with a sensitivity analysis or something like that. Hoping they focus on replicating this value in Slides, Gmail and Docs soon…
2. Website prep via Chrome sidebar
For intro calls, I can open a company’s site and ask with their little Gemini widget on the top of the page:
“I’m speaking to the CX leader, summarize the business model, likely tools, org structure, and draft FPS demand hypotheses.”
Because it knows my workspace and context, this is genuinely helpful.
Future category
Nano Banana image gen might turn into something, but I haven’t tested it enough.
Net: Good for Sheets and browsing. Not viable for writing or any task where tone matters.
Side note - These “industry benchmarks” that claim Gemini and Grok are the best platforms, how do they arrive at that? Am I using them wrong or are these companies paying to get ranked this highly (wouldn’t be surprised) because in my experience they are so much worse than Claude and GPT…
Honorable Mention: Lovable, and why AI (at present) is just a force multiplier
Lovable is an impressive tool and encourage anyone to play with it. You can produce extremely polished, production-level UX without touching code. It excels at creating the front-end of a website or App and giving you a clean design direction.
But without backend or architecture knowledge, or even any coding experience in general, you hit a wall almost immediately. It has integrations with other back-end platforms, but its super hacky and for someone without coding experience is limiting.
This experience showed me that you cannot magically create software without understanding how software works.
This is the clearest example I’ve experienced of the broader reality:
AI is not replacing expertise. AI is amplifying expertise.
Which means:
If you know what you’re doing, AI makes you 2–5x faster.
If you don’t, AI magnifies your blind spots.
The output is only as good as your inputs.
There is no such thing as autonomous AI execution today.
This is why the “AI will replace every job” narrative falls apart when you actually use these tools in real workflows.
What This Year of Daily AI Usage Actually Proved
Great output still requires refinement and a shit-ton of context.
AI has removed a tremendous amount of friction from my day but it has not removed workflows or processes. It has simply increased my leverage.
2026 will be shaped more by integrations than raw model quality. The people who learn to orchestrate tools, documents, systems, and workflows through AI will have an outsized advantage. And I think Claude’s head start here will prove fruitful as the overall efficiency gains will come from significantly increased enterprise adoption.
Teams will run leaner.
Output per person will increase.
Strategy and judgment will matter more, not less.
The disruption will mirror every previous technological shift: uncomfortable at first, but ultimately beneficial.