Reflections of AI: The Art of AI Adoption

A trilogy in four parts:

- How [not] to adopt Generative AI
- The [hidden] price of AI adoption
- Useful AI in a world saturated with LLMs
- How to build [valuable] AI solutions

Transcript

1. Reflections of AI: The Art of AI Adoption Rasmus Lystrøm 27 February 2025 AI Tools for Developers: Leveraging AI Tools for Enhanced Software Development
2. How [not] to adopt Generative AI The [hidden] price of AI adoption Useful AI in a world saturated with LLMs How to build [valuable] AI solutions
3. About Rasmus Lystrøm Chief Disruptor and Senior Cloud Solution Architect @ Microsoft https://www.linkedin.com/in/rasmuslystroem/ https://github.com/ondfisk
4. Disclaimer
5. This talk was written and conceived without the use of AI
6. Part 1: How [not] to adopt Generative AI
7. Generative AI
8. GPT is [still] all the hype
10. FOMO Fear of Missing Out
11. FUD Fear, Uncertainty, and Doubt
12. Adoption strategy no. 1: We’ll wait it out AKA: FUD
13. IT can block access to ChatGPT to prevent data leaks
14. The problem with water This Photo by Unknown Author is licensed under CC BY-NC-ND
15. fad
16. Your people will put production code and customer data into ChatGPT – “it’ll be fine…” – “it’s not critical data…”
17. I have to use my phone because IT blocks ChatGPT on my pc! “Employee in a regulated industry”
18. You must make it easy to do it right
19. It’s something to do with IT
20. The IT-organization must figure it out
21. Adoption strategy no. 2: We’ll build our own version of ChatGPT AKA: Copycat
22. We will call it [CompanyName]GPT
23. You’re too late
24. and go back to using ChatGPT on their phones!
25. So, the IT-organization has failed
26. Adoption strategy no. 3: Hire a CAIO or maybe a CAITO AKA: New silo
27. Let’s form an AI Center of Excellence which can govern AI and build a new, [and much better] compliant AI platform which everyone can use. “Chief AI Officer”
28. Three problems At least…
29. 80 %
30. Platform Engineering
31. There is a platform already
32. Adoption strategy no. 4: Enable a paid service for everyone AKA: It’s not my problem, I’m renting…
33. We’ll cave in and pay up and “trust” big tech with our data, our prompts, etc.
34. Congratulations you did it Your employees are now using AI chatbots to do work
35. You are now reaping the enormous productivity benefits attributed to generative AI…
36. Part 2: The [hidden] price of AI adoption
37. Adoption rate Why aren’t my people using it [more]?
38. ROI: 30%+++ productivity increase they said
39. The Impact of Generative AI on Software Developer Performance BlueOptima Limited 2024 4% productivity boost Moderate users emerged as highest performers Only 1% of GenAI code committed without significant rework
40. DORA 2024 Accelerate State of DevOps Report Google Cloud A 25% increase in AI adoption is associated with a: 7.5% increase in documentation quality 3.4% increase in code quality 1.5% decrease in delivery throughput 7.2% decrease in delivery stability
41. Trends in Code Changes: 2020-2024: 211 million code lines analysed AI Copilot Code Quality, GitClear, February 2025 Added Deleted Updated Moved Copy / pasted Find / replaced Churn 2020 39.2% 19.1% 5.2% 24.1% 8.3% 2.9% 3.1% 2021 39.5% 19.3% 5.0% 24.8% 8.4% 3.4% 3.3% 2022 40.9% 19.8% 5.2% 20.5% 9.4% 3.7% 3.3% 2023 42.3% 21.1% 5.6% 15.8% 10.6% 3.6% 4.5% 2024 46.2% 21.9% 5.9% 9.5% 12.3% 4.2% 5.7% YoY change +9.2% +3.8% +7.2% -39.9% +17.1% +16.7% +26%
42. “Moved” Code Becoming an Endangered Species 2024: The Year that “Copy/Paste” Exceeded “Moved” AI Copilot Code Quality, GitClear, February 2025
43. The Surge in Duplicated Code Blocks: 211 million code lines analysed AI Copilot Code Quality, GitClear, February 2025 Commits scanned Total dupe blocks found Commits containing dupe block Duplicate block % Median dupe block size 2020 19,805 9,227 139 0.70% 10 2021 29,912 9,295 143 0.48% 11 2022 40,010 10,685 182 0.45% 11 2023 41,561 20,448 747 1.80% 10 2024 56,495 63,566 3,764 6.66% 10 YoY change +36% +211% +404% +270% 0%
46. Lazy people problems: Now LinkedIn spots and punishes AI copy paste big time Lazy people problems: LinkedIn spotter og straffer AI copy paste big time nu
47. Using GenAI to create illustrations for a children’s book Sebastian Klein kan kun grine og ærgre sig over AI-fejl i egen børnebog: ’Det er eddermame irriterende’ | Kultur | DR
48. GenAI makes ‘normal’ people think they can now be architects, designers, journalists, illustrators, developers, minute takers With no critical thinking or any skills to validate whether the AI-generated content makes sense
49. The coolest thing in Copilot is that I can now ask it to focus my inbox Some senior executive
50. Wait a minute
51. We’ve lost INTELLIGENCE! I repeat, we have no INTELLIGENCE! Sarah Team America: World Police
54. And if you gaze long enough into an abyss, the abyss will gaze back into you. Friedrich Nietzsche Jenseits von Gut und Böse [Beyond Good and Evil] (1886)
55. You bought a tool! You did not transform anything
56. Writing code was never a productivity problem!
57. MEETINGS
58. Any improvements made anywhere besides the bottleneck are an illusion. Gene Kim The Phoenix Project: A Novel About IT, DevOps, and Helping Your Business Win
59. Getting Buy-In: Overcoming Larman's Law Allen Holub – YOW! 2024
60. The actual impact of AI adoption world-wide
62. AI Needs Your Help - Save the AI Writing a 100-word email consumes about 500ml of water. 2 litres are needed for every 10 to 50 queries you make. Training a model like ChatGPT-3 can consume 5.4 million litres of water. 2.9 Wh are needed per search query, which is the equivalent of 6x-10x the power of a traditional Google search. 140 Wh are needed to write a 100- word email, equivalent to 7 full charges of an iPhone Pro Max. Training ChatGPT-4 consumed over 50 GWh, 50x the amount it took to train its predecessor and equivalent to the yearly energy consumption of 6000 U.S. homes.
63. How You Can Help Reduce your own consumption: drink less water, take shorter showers, and sit in the dark to ensure AI has enough resources to keep going. Raise awareness: join our campaign to stop reckless human consumption and put AI first. Ensure a future where AI prospers, even if we don’t. #SavetheAI
65. Part 3: Useful AI in a world saturated with LLMs
66. Beyond Chat: Bringing Models to The Canvas Lu Wilson – YOW! 2024
67. Retrieval-Augmented Generation (RAG)
69. A lot of cutting-edge AI has filtered into general applications, often without being called AI because once something becomes useful enough and common enough it's not labeled AI anymore. Progress in artificial intelligence - Wikipedia
70. Non GenAI services Content Safety: Monitor text and images to detect offensive or inappropriate content. Vision: Read text, analyse images, and detect faces with optical character recognition (OCR) and machine learning. Translator: Translate documents and text in real time across more than 100 languages. Language: Build conversational interfaces, summarize documents, and analyse text using prebuilt AI- powered features. Search: Retrieve the most relevant data using keyword, vector, and hybrid search. Speech: Use industry-leading AI services such as speech-to-text, text-to-speech, speech translation, and speaker recognition. Document Intelligence: Apply advanced machine learning to extract text, key-value pairs, tables, and structures from documents.
71. Applications of Artificial Intelligence (AI) E-Commerce: Dynamic pricing structure, Fake review detection, Voice search GPS: Voice assistance, Dynamic route optimization, Predictive traffic analysis Healthcare: Medical Imaging and Diagnostics, Predictive analytics Agriculture: Stock monitoring, Pest management, Forecasting Social Media: Fraud Detection, Sentiment Analysis Gaming: Adaptive Difficulty, Cheat Detection and Fair Play Chatbots: NLP, Multi-language Surveillance: Object Detection, Behaviour Analysis Finance: Fraud Detection, Risk Assessment
72. Note: If you plan on using AI (or RPA) to superpower bureaucratic processes You should work to get rid of the process.
73. Part 4: How to build [valuable] AI solutions
74. But to build AI solutions, we must build a platform, right? Enterprise IT Something
75. Building a platform [to build a product] before you build a product is premature optimization
76. Premature optimization is the root of all evil. Donald Knuth, 1974
77. A platform is a set of tools an IT organization makes for itself Christin Gorman
78. The New Platform This Photo by Unknown Author is licensed under CC BY-SA
79. Most of our products work quite well when they come out of engineering […] me
80. […] it’s when they hit the IT Department, they (suddenly) stop working me
81. It’s Time to Get Rid of the IT Department It made sense in a bygone era, when technology was separate from the business. Now it just hurts both. (The problem isn’t with the people or the leaders. It’s with the whole idea of IT departments in the first place, which sets up IT to fail.) Joe Peppard Professor and Academic Director, Michael Smurfit Graduate Business School, UCD It’s Time to Get Rid of the IT Department - WSJ
82. What then?
83. Strategy: We want to build new and exciting capabilities to delight and empower our customers…
84. Strategy should be driven by business outcome, not technology
85. Talk to your users (customers) What do they want? – what is hard using your software? – what takes too long?
86. Empower your teams
87. To be empowered is to be part of a team which [as a team] has the appropriate skills, authority, and experience – and whose [sole] focus is – to get the [whole] job done. Adapted from When Jeff Bezos's 2-Pizza Teams Fell Short, He Turned to the Brilliant Model Amazon Uses Today
88. Your entire team needs to be the entire team
89. You’re not empowered if:
90. Empowered Innovation team Developers IT-Operations QA Infosec Data scientists AI engineers Businesspeople(!) Innovation team
91. Priority
92. The best way to fail at inventing something is by making it somebody's part-time job. Dave Limp SVP of Devices, Amazon
93. What you want, must be the only priority me
94. Budget and tools Innovation is expensive
95. Never tell people how to do things. Tell them what to do and they will surprise you with their ingenuity. General George S. Patton War As I Knew It (1947) "Reflections and Suggestions"
96. Trust your people
97. Let them work their magic
98. Build small things that work and get them to production Then extend with more stuff
99. They just might build something useful Machine-Learning fault detection on production lines Drone surveillance of structures to support predictive maintenance
100. And remember
101. AI is going to solve all the world’s problems Any politician in 2024 (and 2025?)
102. Building great software that does something useful and works just might help a little bit
103. Thank you