While we have all witnessed highs and lows of AI, accelerated and fueled by GenAI in the last 2 years, where everyone in every industry including IT would assume You are stupid if you can't spell that out, yet there are very little concrete evidences put forward by business and technologies companies to fortify their claims and how fantastically they have been solving the problems which were not possible earlier at all or how they have taken a very transformation approach to solve them beside everyone in competition is left behind and hence "You should use our products and services", if you want to join that "elite group" as well.
Now there are perhaps enough talks, materials, white papers and podcasts which you might have come across at some point supporting that hype.
On the contrary if You speak with the real world practitioners (hard to identify and find) or me who don't easily get intimidated by the "shiny new tech", here are few things You may want to consider while picking up your:
- Preferred Solutions
- Preferred OEMs
- Preferred Systems Integrator
- Preferred Managed Services Provider
Why Should I care ? - Well actually you should, See - If you are still looking for answers to these questions, You have likely missed the bus and is now trying to cope up with the industry and your competition. Or perhaps, it's a mandate given by business to your company's IT team, Product team and so forth.
Now you might have been playing with some of Gen AI capabilities either for fun or really seriously (outside mandates from business) using the public clouds, since its cheaper to play, test and throw/dismantle and do it all relatively cheaply, but in the absence of a firm "Strategy", you have either "underdelivered" or "failed" or perhaps "had little success" and want to leverage those learnings to further fortify your strategy and make course corrections.
Meanwhile the big management and IT consulting firms must have been reaching out to your Board, MD & CXOs claiming that they have successfully solved this puzzle and have enough evidences and credibility to showcase and telling you that you should now onboard them as partner to this endeavor into unchartered territories. It will only cost You few millions $$$ and 6-12 months of time to transform your business forever, And how you will be the next "MAANGx" - X being the latest addition to the group as You. Happy Ending !!!
What you either often forget or have a "dejavu" moment of is:
I have seen everything working perfectly on presentations earlier too beside from my partners - consulting firms, vendors, large public cloud providers, systems integrators or managed service provider telling me how they are putting their credibility and life on the line for me to be successful.
So remember, as an enterprise (small-mid-large) You have:
- A large baggage of your technical debt
- Absence of strong and foundational architectural + design practices within IT, though on paper everyone claims they have
- Believing this time Your partners in crime won't ditch you or would leave you in the middle
- No incentives to put together a good solution but a solution that serves the immediate needs/mandates, it always works like that in Enterprises and Telcos, later you can blame it on - pace of technology change, people issues, process issues, technology issues, urgent needs etc...
- Hiring smart people only solve half of the problem at best ...ever
So rather ask your "Partners" to move beyond the presentations, market research data and help you with:
- A short-mid terms AI strategy
- What accelerator's, assets, frameworks, reference architectures, models, test results they have in place to expedite the whole process with speed and agility and yet minimizing the risks
- Show, don't tell about - Your experiences (Industries, scale, labs, poc/pilot, expertise & depth, partner ecosystems with level of innovation/MVPs, success stories that you can validate yourself beyond the presentations
- Help build a strong and evidence based business case with ROI instead of just a bunch of excel sheets thrown on you
- Can they do a BTO (Build, Transfer & Operate), Org. wide AI upskilling is a long uphill battle which most are loosing without admitting at this point
- Choice of consumption models (IaaS & PaaS), don't pour all your money into promises
And at last, the technical details around:
- Clear cut business cases
- Supported models (small, large, multi-layer, multi-tier)
- Accuracy (data) & course correction methods + Architectural tradeoffs
- Supported vs. required integrations
- Quality and controls
- Runtime environments
- Performance boost methods, metrics and measurements
- Dev vs. Test vs. Production pipeline
- Tokens (Free vs. Paid, How many, Cost per token, Cost breakdown per token use, Length of Prompt/Description, Error rates)
- Hallucinations
- Privacy, compliances and regulations
- Sustainability footprint (ESG metrics and Power needs)
- Investment protection
- High level and tentative cost + efforts for each level of customizations
- Adoption roadmap, milestones, success metrics and RACI
- Governance & risks mgmt.
- Cost management, quality assurance & ROI
- Program structure & stakeholders
And next time when you read about - how "X/Twitter" has put together a data center running 100K+ GPUs in just 19 Days - Remember you are not "Elon Musk" and most likely your company don't have one either beside the risk appetite that "X" has is far bigger.
But otherwise, all You know are the execution details, behind the scenes how much time and effort they have put in for coming up with a strategy and plan is what you no clue about, beside they are known for they high standards and hence they became this big.
But if you think "strategy" is something you can fully "outsource", you are bound to learn your lessons sooner than later.
So stay grounded and keep it real.
HTH...
A Tech Artist 🎨
Further Readings
Nvidia eyes data center Ethernet as its next multi-billion-dollar biz
Incorporating generative AI into your company’s technology strategy
Amazon, Google make dueling nuclear investments to power data centers with clean energy
AI and its carbon footprint: How much water does ChatGPT consume?
Meta's Mark Zuckerberg says energy constraints are holding back AI data center buildout
Why Zuckerberg’s multibillion-dollar gamble doesn’t just matter to Meta
Has the AI bubble burst? Wall Street wonders if artificial intelligence will ever make money
The AI bubble has burst. Here's how we know
Bonus Materials For GenAI Enthusiasts From Web - Credit to Original Creators
No comments:
Post a Comment