{"id":51,"date":"2025-04-17T18:10:14","date_gmt":"2025-04-17T16:10:14","guid":{"rendered":"https:\/\/www.romainpechard.com\/?p=51"},"modified":"2025-04-17T18:10:14","modified_gmt":"2025-04-17T16:10:14","slug":"spending-3-months-full-time-working-with-chatgpt-lessons-learned","status":"publish","type":"post","link":"https:\/\/www.romainpechard.com\/index.php\/2025\/04\/17\/spending-3-months-full-time-working-with-chatgpt-lessons-learned\/","title":{"rendered":"Spending 3 months full time working with ChatGPT: lessons learned"},"content":{"rendered":"\n

The invisible team member<\/strong><\/p>\n\n\n\n

ChatGPT is the greatest fulltime, always-on personal assistant I\u2019ve ever worked with. It operates at the level of a confirmed senior consultant. It can handle and properly use advanced frameworks \u2014 RAPID, 3 horizons, PESTEL, Creative Thinking, Business Model Canvas, TOM Canvas, Value Proposition Canvas, and plenty more \u2014 without flinching. The catch? You need to know how to talk to it.<\/p>\n\n\n\n

Prompting is a craft. Either you frontload everything in an initial prompt, injecting your methods and tools as defaults, or you structure your thinking with care along the way. Precision matters. Sequence matters. And the framing? That\u2019s half the job. Like coding \u2014 nothing new there \u2014 but it forces a level of discipline. Lose it, and you\u2019ll get blunders, inconsistencies, or worse: something that sounds smart but is completely off.<\/p>\n\n\n\n

Bend it, shape it, rewire it<\/strong><\/p>\n\n\n\n

The second thing that stands out after three months is how easily you can bend it to your will \u2014 and how absurd it can become if you’re not paying attention. It will happily change its mind if you tell it its reasoning is wrong. That\u2019s not a bug, it\u2019s a feature.<\/p>\n\n\n\n

The cheat code? Ask it to discuss its own reasoning. Tell it to simulate multiple points of view \u2014 profiles you\u2019ve defined in your initial prompt \u2014 and it will adapt its logic accordingly. That first prompt is critical. It lays the ground for how it responds, how it argues, and what “style of thinking” it defaults to. Set it right, and it becomes a much sharper tool.<\/p>\n\n\n\n

Fake it like agentic AI<\/strong><\/p>\n\n\n\n

Right now, agentic AI is trending. Fair. But you can already simulate that with ChatGPT \u2014 if you take on the agentic part yourself. Build a structure. Don\u2019t ask it to answer. Ask it to list possible angles. Then ask it to build a document out of those. Then expand each option. One by one. Kill bad paths, go back if needed, branch out if new info surfaces.<\/p>\n\n\n\n

This way of working rewards tree-like thinking. If you’re wired like that, it clicks fast. If not, no problem \u2014 just have it write you a reminder and follow that method. Personally, I didn\u2019t bother embedding the process into my initial prompt. I know how to run it. But it can work either way.<\/p>\n\n\n\n

don\u2019t think straight<\/strong><\/p>\n\n\n\n

I know some will default to linear thinking. That\u2019s a valid approach, and it can work \u2014 to a point. But if you\u2019re not exploring side paths, if you\u2019re not simulating subject matter experts in a room, brainstorming from different angles, you\u2019re missing what GPTs are built for. You\u2019re flattening multi-agent potential into a narrow Q&A box. That\u2019s not what this is for.<\/p>\n\n\n\n

You want a metaphor? Call it controlled schizophrenia.<\/p>\n\n\n\n

How fast can you think?<\/strong><\/p>\n\n\n\n

What else? Framework building, idea consolidation, capability projection, business planning \u2014 even for complex topics like strategic foresight, revenue growth management, business modeling \u2014 all become ridiculously fast. Not perfect, but fast and structured.<\/p>\n\n\n\n

What would take you two or three weeks with experts can be mapped in minutes. You won\u2019t land on a polished output \u2014 you\u2019re not shipping this as is \u2014 but you will have identified all the dimensions. You\u2019ll have a structured overview in 15 minutes, a decision-ready deep dive in an hour.<\/p>\n\n\n\n

Then just wrap it in Word or PowerPoint \u2014 ChatGPT is terrible at that part, don\u2019t even try \u2014 and you\u2019re good to go. Add tailoring, audience calibration, and you\u2019re ready to pitch.<\/p>\n\n\n\n

Comparing notes with a team<\/strong><\/p>\n\n\n\n

Let\u2019s be honest: ChatGPT, used right, is like working with a consulting team of five or six. Four juniors and two managers, maybe. It doesn\u2019t replace even one of them individually \u2014 it can\u2019t run solo. You can\u2019t leave it alone, check back later, and expect gold. It needs ongoing pressure, iteration, framing. But the output \u2014 if you stay sharp and drive the process \u2014 is comparable to what that team would deliver together.<\/p>\n\n\n\n

You need to stay in control, challenge the dead angles, pressure test the logic, ask for opposing views, factor in unknown unknowns. It requires backward thinking, layered reasoning, on-the-fly research, and yes \u2014 humility.<\/p>\n\n\n\n

Consulting, scaled like SaaS<\/strong><\/p>\n\n\n\n

But if you\u2019ve got that mindset \u2014 if you know how to run a P&L and you understand that the Business Model Canvas is a cheat code to unfold every strategic layer \u2014 then this becomes your multiplier. You can turn one team member into a six-person A-Team. You can scale your consulting work like SaaS scaled software.<\/p>\n\n\n\n

It\u2019s not about replacing expertise. It\u2019s about giving that expertise exponential leverage.<\/p>\n\n\n\n

That\u2019s the game.<\/p>\n","protected":false},"excerpt":{"rendered":"

The invisible team member ChatGPT is the greatest fulltime, always-on personal assistant I\u2019ve ever worked with. It operates at the level of a confirmed senior consultant. It can handle and properly use advanced frameworks \u2014 RAPID, 3 horizons, PESTEL, Creative Thinking, Business Model Canvas, TOM Canvas, Value Proposition Canvas, and plenty more \u2014 without flinching. […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[36],"tags":[61,63,60,62,64],"class_list":["post-51","post","type-post","status-publish","format-standard","hentry","category-ai-autonomy-performance","tag-agentic-ai-workflows","tag-ai-augmented-strategy","tag-generative-ai-in-consulting","tag-prompt-engineering","tag-scaling-consulting-with-ai"],"_links":{"self":[{"href":"https:\/\/www.romainpechard.com\/index.php\/wp-json\/wp\/v2\/posts\/51","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.romainpechard.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.romainpechard.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.romainpechard.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.romainpechard.com\/index.php\/wp-json\/wp\/v2\/comments?post=51"}],"version-history":[{"count":1,"href":"https:\/\/www.romainpechard.com\/index.php\/wp-json\/wp\/v2\/posts\/51\/revisions"}],"predecessor-version":[{"id":52,"href":"https:\/\/www.romainpechard.com\/index.php\/wp-json\/wp\/v2\/posts\/51\/revisions\/52"}],"wp:attachment":[{"href":"https:\/\/www.romainpechard.com\/index.php\/wp-json\/wp\/v2\/media?parent=51"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.romainpechard.com\/index.php\/wp-json\/wp\/v2\/categories?post=51"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.romainpechard.com\/index.php\/wp-json\/wp\/v2\/tags?post=51"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}