A4) As [Ali] said, the experience gained with public #cloud's variable costs has been valuable when accounting for token burn. So far, tracking closely ane optimizing at strategy / design time where we can. #ciochat
1
0
0
0
Hashtag
#CIOChat
Advertisement · 728 × 90
A3) So far, working that into project and product budgets and forecasts. As [Isaac] said, those forecasts have to stay pretty darn loose since estimation is rough at best. Model churn doesn't help. #ciochat
0
0
0
0
A2) Prompt discipline / design and model choice more than in-line model routing. Maybe a path to fewer surprises, maybe not. #ciochat
0
0
0
0
A1) Not ramping anything new right now, so [token burn is] flat-ish with existing people and services. Definitely not in the "track employee token burn as a proxy metric for their value creation" camp. #ciochat
0
0
0
0
A4) All of the new capabilities you mentioned earlier, but I am fatalistic about the risk. We have to have (and maybe steer) open, honest, clear-eyed talks at high levels to manage the speed / risk trade-off and ensure alignment with share- / stakeholder risk appetite. #ciochat
1
0
0
0
A3) Decentralized ownership on top of reusable architectures to enable the business folks to own the things that might normally turn into #ShadowAI. Sadly, an #AI center of excellence scales even worse now than relying on one to keep #cloud in check. #ciochat
0
1
0
0
A2) Whenever possible, making a "plaform play" and defining reusable components, infrastructures, and architectures that we can use to speed up the rest (and accelerate the cycle). Complexity and super-sprawl (much like microservices) are two of the enemies. #ciochat
0
0
0
0
A1) Stakeholder alignment, but that's partly because I insist on security, audit, privacy, and risk folks being first-class stakeholders and brought in from day zero whenever possible. #ciochat
0
0
0
0
A4) If the #CIO doesn't bring more value than that, they'll be replaced by the CIObot (w/ maybe a fractional/virtual human in the loop) in no time. I'm not going to do the McKinsey thing and count bots as "digital employees". #ciochat
0
0
0
0
A3) That is the big Q! One question I ask everyone is "are you building the simolest solution that fills the need?" Sprawl (of every kind) and complexity are the enemies of predictability going in and transparency / explainability coming out. #ciochat
0
0
0
0
A2) 😆😢😆😢😆😢 Without making the teams feel threatened? That is, sadly, not the story of anywhere I've been lately. Whether it's the algorithms, the out-sourcing, the downsizing, or the M&A; the pressure is building. #ciochat
0
0
0
0
A1) Nothing too sexy here. The usual kinds of ticket routing, workflow, and #SLA management were the start. #FinOps brought a lot of data requirements and algorithms in as well. #ciochat
0
0
0
0
A4) Sounds like a good recipe. Aligning w/ business goals and keeping one eye outside the company on customers competitors, potential disruptors, and opportunities is becoming a #CIO necessity (almost) w/o regard to the make-up of the C-suite. #ciochat
0
0
0
0
A3.2) Literally *everything* is connected, and pretending otherwise just gets us in trouble. Holistic approaches are harder and sometimes costlier up front, but finding and debugging the issues between services (at every level) reduces risk across the board. #ciochat
0
0
0
0
A3.1) Balancing "left of boom" and "right of boom" planning & prep. Tabletop exercises to keep costs down but vigilence up. The real difference between #IT Service Continuity and true business continuity... #ciochat
0
0
1
0
A2) Prioritizing relationship management, partnerships across the C-suite and LoBs, and proactive, transparent communication so nobody has to guess when / why / whether we're super busy or running short on other resources. Not redefining success criteria midstream. #ciochat
0
0
0
0
A1) At the last day job, it was the dark side of "ownership". Not just petty fiefdoms and bickering but "we own this [data, service, contract, whatever] and can cut everyone out as much as we want". The "... and find out" stages were predictably brutal. #ciochat
0
0
0
0
A4) Continuous #FinOps. Predict what we can with relative confidence but track everything in real time that we can. #ciochat
0
0
0
0
A3) Not sure I'd go so far as to say "probabalistic models", but wrapping some explicit confidence intervals around all the raw numbers provides some level of reputational armor. #ciochat
0
0
0
0
A2) Yes, and leveraging as much #FinOps as possible to understand current / historical costs and help us predict as much as can within a decent confidence interval. #ciochat
0
0
0
0
A1.3) Regardless or breakdown or percentage, an org's #innovation budget probably shouldn't have been subject to hard caps unless absolutely necessary. #ciochat
0
0
0
0
A1.2) There are still some foreseeable and controllable costs (cloud or otherwise), but the CapEx-or-bust crowd has had a major impact there. #ciochat
0
0
1
0
A1.1) Looking back over a few orgs, that started long before the current #AI hype wave for sure. All the rest have played roles along with virtualization, a drive for self-service infrastructure provisioning (cloud or not), etc. #ciochat
0
0
1
0
A4.2) All of the above in part because pulling back from some of the earlier decisions carries costs for which we're not ready. Double down on the learning & understanding bits of your org. None of this works if you're operations-only and hoping providers do it all. #ciochat
1
0
0
0
A4.1) Cloud-smart but likely with an edge toward cloud-first in many orgs because everybody is still under pressure to optimize spend and focus on revenue generation, business opportunity enablement, etc. #ciochat
1
0
1
0
A3) I'm feeling lucky at the moment not to be in a multinational or a government-facing entity that needs FedRAMP / GovCloud. But, recent moves in the EU do seem to signal another bloc pulling away from US #cloud hegemony. #ciochat
1
0
0
0
A2) Not trying to compete in #AI model building, primary training, etc. gives more options for where to run the resulting inference workloads. The hardware and software options for that are all over the map from GPUs to baked in accelerators to #IBM #Spyre and more. #ciochat
0
0
0
0
A1.2) Sadly, that stemmed from every possible bad fit. Too little move to the #cloud? Problems! Too much? Problems! Wrong stuff or wrong order of moves? Problems! Public cloud success was always going to be a Goldilocks story. #ciochat
1
0
0
0
A1.1) As a "right tool for the job" person, I never went all-in on OpEx over CapEx #SaaS, or #cloud. All have big strengths, but all have big caveats. Not sure the discussion is still "quiet" about cost & complexity spirals and either repatriation or shuffles. #ciochat
2
0
1
0
A4.1) The recent news items about Amazon and others "tightening up" their review processes are good examples of environments where you cannot wait for the review to find out how you're doing. Set & chart your own goals and wins. #ciochat
0
0
0
0