Hiring for AI skills + GPT-5

AI Skills Assessment | GPT-5's Coding Focus | GPT-5 Writing Example | GPT-5 by Skill Levels | Vibe coding with GPT-5 |. Personal GPT-5 Recommendations | Gemini Drive Integration

If you only check one highlight this week:

  • Team leader? Hiring for AI skills

  • Everybody else: meet GPT-5 your new day-to-day AI partner

-FranƧois

Thanks for sharing these highlights with busy marketing execs around you. šŸ™ 

Someone forwarded you this email? You can subscribe here.

Why (and how) I'm now assessing AI skills in EVERY key hire

Over the last few months, I've helped clients scope and assess candidates for eight senior roles. For every single one, AI proficiency ended up being a ā€œmust-possessā€ on the scorecard, never a nice-to-have (bear in mind that all my clients are in B2B tech, with most of them in AI and/or developer tools, so the bar may be higher but I believe this will quickly spread everywhere).

Why I want AI-savvy hires

If you're reading this, you know why. But it's good to remind ourselves so we don’t short change that part of the assessment and dedicate at least half of a focus interview to the topic:

1ļøāƒ£ Employees who already use AI well are typically self-learners (there's little up to date corporate training for AI yet). That signals a growth mindset, an eagerness to scale, and the kind of agility and adaptability teams need today, when so much is changing fast.

2ļøāƒ£ AI-fluent employees can not only do more work, they can do more strategic work. They use AI to research, plan, analyze, decide and then create, prototype, and automate repetitive tasks. They focus on higher-value activities while getting more done overall.

3ļøāƒ£ They become force multipliers. The best AI users naturally help others on their team level up. They share what works, create templates, and elevate everyone's productivity.

How to actually assess this

I ask three specific questions:

1ļøāƒ£ "Show me your three most advanced uses of AI at work. Why and how did you build them?"
Ā» Look for examples that teach you something new. Strong candidates identify the right workflows to augment, use Projects, build custom GPTs, use automation tools, vibe code apps and combine different tools.

2ļøāƒ£ "How have you helped others on your team use AI or automation? What recognition did you get?"
Ā» You want people who evangelize and get pulled into teaching others, not just solo operators.

3ļøāƒ£ "What are your favorite tools. What resources do you follow and recommend to keep up with new capabilities and tools and why?ā€
Ā» Assess familiarity with tools beyond ChatGPT. See if they follow intermediate or advanced influencers or resources plus glean a few you may want to follow to. (if they read my highlights = HIRE the person on the spot!!! ;-))

What about you?

Try answering these questions yourself and see how you’re doing. If you're hiring, I recommend you go for candidates that do even better than you did.

Of course, by asking these questions of candidates you will be learning a lot during the interviews! If you don’t learn anything new, kindly pass.

My selection of tips, news and workflows

If you didn’t have time to read about or push GPT-5 šŸ‘‡

šŸ’» GPT-5 Part I - An upgrade focused on coding

Most anticipated, slightly disappointing to many because it seems incremental progress at first (except for coding), the rollout of GPT-5 was the anticipated news of the week.

Where it impresses:

  • Works and thinks harder longer if it needs to (goodbye confusing model picker*).

  • Can use multiple tools simultaneously with high accuracy

  • Focused on the coding use case (big and fast-growing market and use-case): nicer designs, able to create more complex apps, debugs better, less expensive than Claude Code via API.

  • Faster and more reliable

Some don't believe it's a great improvement for writing (but check out the example by Ethan Mollick below and be your own judge...).

Reid Hoffman focused his take on ā€œOpenAI’s strategic decision to deprecate all previous models in favor of GPT-5 alone, with various fine-tuned versions running quietly in the backgroundā€, since AI will be simpler to use for most people.

šŸ“ GPT-5 Part II - This is impressive but NOT the main upgrade

Professor Ethan Mollick was one of many people having early access. He shared this example in his ā€œIt just does stuffā€ review. Pay attention to what's under the result.

Prompt:

ā€œYou are GPT-5, do something very dramatic to illustrate my point (that ā€œthis is a big dealā€), it has to fit into the next paragraph.ā€

It ā€œthoughtā€ for 24 seconds and wrote:

ā€œThunderstruck. Here, watch. I build worlds. See ideas become instruments. I code, compose, and converse. Stories synthesize science; solutions spark, swiftly. Asks arrive; answers appear, astonishingly, across domains. Behold: business plans, blueprints, briefs, beautifully built, briskly. I infer intent, integrate insight, and iterate instantly, indefatigably. Gaps get gauged; goals get generated, guarded, and guaranteed gracefully. During dialogue, disparate data distills down; decisions develop deliberately, deftly, dependably. Everything explained efficiently, even ethics: evidence examined, empathy engaged, everywhere, everyday, equitably. All at once, ambiguous arguments align as actionable, accurate advice awaits anyone anywhere. Look: language, logic, learning, and long-term memory link, lifting lives and labor limitlessly worldwide.ā€

GPT-5’s answer to Mollick’s prompt

He then reports:
ā€œIf you didn’t catch the many tricks:

  • the first word of each sentence spells out the phrase This is a Big Deal

  • each sentence is precisely one word longer than the previous sentence.

  • each word in a sentence mostly starts with the same letter, and it is coherent writing with an interesting sense of style.

In a paragraph, GPT-5 shows it can come up with a clever idea, plan, and manage the complicated execution (remember when AI couldn’t count the number of Rs in ā€œstrawberryā€? that was eight months ago).ā€

The example above is nothing compared to the apps that some folks report this model can help build.

šŸ“ˆ GPT-5 Part III - What it improves by current skill level

A recap by Nicole Leffer on what you should benefit from

šŸ‘‰ FOR BEGINNERS
ChatGPT will probably seem smarter.

Sometimes it will switch (on its own) into "thinking" mode, and when it does that it will seem a LOT smarter than you're used to. If you definitely want it to think deeply about something before answering, you can tell it to in your prompt.

If you use a ton of GPT-5 in any 3 hours, it may switch to less smart version of the model until those 3 hours are up. (This happens in Free and Plus only)

šŸ‘‰ FOR INTERMEDIATE USERS

You'll now only have two model choices (unless you're in Pro): GPT-5 and GPT-5-Thinking.

- GPT-5 is a slightly smarter equivalent to GPT-4o.

- GPT-5-Thinking is a slightly smarter equivalent to o3, with faster reasoning.

If you're in GPT-5, sometimes it will automatically switch to reasoning (i.e. "Thinking") to respond.

If you have GPTs, they'll all run on the GPT-5 family now, too.

There will be less hallucinating (especially for Thinking vs. o3), it won't agree with you when it shouldn't as much, and it will follow instructions better.

šŸ‘‰ FOR ADVANCED USERS

It's important to understand GPT-5 isn't actually a single model. It's a system of multiple models, some non-reasoning, some reasoning, and a router that chooses the best model.

It will look for explicit intent in your prompt to help select the best model and to determine how much reasoning to do (if any). (Tip: your prompt can guide this!)

If you're using GPT-5-Pro (currently in Pro accounts), the model launches several lines of reasoning at the same time and then picks or combines the best result. It’s slower, but should be better on hard problems.

The models are great at instruction following.

Poorly constructed prompts can degrade performance - especially if instructions contradict. Make sure to explicitly prompt how to prioritize conflicting instructions.

GPT-5 (especially 5-Thinking) has strong planning and self-reflection abilities, and you can prompt for that.

Thinking is also really good at coding - I experimented with single shot website landing page development in Canvas and was stunned how good it did. (Should also be great at creating things like apps you can put on your website!).
ChatGPT will probably seem smarter.

Nicole Leffer

šŸ‘Øā€šŸ’» GPT-5 Part IV - My own vibecoding test

I like uploading my content into a vibecoding tools (Claude Artifacts, Lovable, etc.) and then starting with a loose prompt to see what it’s capable of.

I ask it to create an interactive app based on my content. I mention website styles that I like and see a first pass

This is what I vibecoded with GPT-5 in five shots only based on questions that I had crafted for CEOs and CMOs can assess their marketing game and alignment.

It's still a very early prototype, but pay attention to the effects, graphics and animations (he 12 questions were written by me, ChatGPT generated the rest of the content and it’s not there yet).

šŸ’” GPT-5 Part V - What it means for me (and you)

I then simply asked GPT-5 to tell me, given my usage history, how I will benefit from GPT-5 compared to o3, my favorite model.

You can try the same prompt and get your own personal recommendation:

Based on what you know of my profile and history, how I use ChatGPT for work, and the context I've given you, give me a very tight summary of the benefits of the new capabilities of GPT-5 vs o3 and recommend the top 5 use cases I should be using it for that are now possible or greatly improved.

It answered:

ā€œWhat you get over o3

  • Unified ā€œsmart-vs-thinkā€ system + routing. GPT-5 auto-decides when to answer fast vs engage deeper reasoning. ļæ¼

  • Higher real-world accuracy & fewer hallucinations with better instruction-following and open-ended factual checks. ļæ¼

  • Stronger coding + agentic tool use.

  • Better multimodal performance (text + visuals). ļæ¼

For context on o3: it was the prior top reasoning model. Great at complex analysis and visual tasks, but without GPT-5’s routing system and broader ā€œdefault modelā€ role. ļæ¼ ļæ¼ ļæ¼

Your top 5 upgraded use cases (vs o3)

  1. ā€œOne-brief-in, plan-outā€: Give GPT-5 a messy launch brief and say ā€œthink hard.ā€ It will route to deeper reasoning, keep constraints straight, and return a tightened GTM plan with fewer follow-ups than o3. ļæ¼

  2. Messaging house generation that actually sticks to your voice across long sessions—less prompt-rehydration, tighter instruction following. ļæ¼

  3. Competitive intel → POV in one pass. Ask for multi-source synthesis + recs; GPT-5’s improved factuality and ā€œagenticā€ tool use reduces manual cleanup. ļæ¼

  4. Interactive asset prototyping (calculators, assessments, mini-apps). GPT-5’s coding wins + agentic patterns mean faster working prototypes. ļæ¼

  5. AIO/LLM-SEO monitoring loop: Have it routinely analyze AI search outputs and suggest content fixes—benefits from GPT-5’s accuracy + routing for longer reasoning cycles.ā€ ļæ¼

✨✨ Gemini in your Google Drive and Files (because there's more to life than ChatGPT…)

If you don’t work for a large company, then most of your files may live in Google Workspace and Drive, like me. Gemini is able to do more and more with that and it's a massive time saver, given it’s right here.

  1. With a ā€œMarketing Planā€ Google Doc open, I asked Gemini to critique the plan without holding any punches. From structure to content. I was impressed by the speed and relevance of the answer. Of course it is not meant to replace your judgment, but it is a good thought starter. I also really like summarizing contracts, asking for their analysis and areas I should pay attention to + tips to negotiate.

  2. In Google Sheets, the ā€œ=AIā€ function allows you to summarize or categorize content from other cells, or simply generate. See screenshot below.

  3. In Google Drive, you can have Gemini summarize entire folders. You can ask questions and find files faster.

Final Words

ā

GPT-5 is a tangible example of the coming abundance of intelligence. It's in your pocket. It meets you and your problems where you are. Wherever in the world you are, in whatever language you speak, at whatever level you want it to, with limitless patience.

Reid Hoffman

FranƧois | LinkedIn 

I'm a CMO, advisor, and "CMO Wingman". Yes, that's a thing :-). Ask the CMOs I support: in this AI era, CMOs need a strategic proactive advisor more than ever. I’m former CMO at Twilio, Augment Code, Apollo GraphQL, Decibel, Udacity and Head of Marketing for LinkedIn Talent Solutions.