eWeekChat

A4: An audit can help pre-determine what a successful #AIOps implementation looks like for an organization. This includes aligning your AIOps initiative to address current needs such as more effective event noise reduction & faster probable cause analysis.

A4: Start with the fundamental. What business problem are you trying to solve and why AI. Some business challenges may not require sophisticated AI models.

A4: This is a little counterintuitive but I would say don’t focus on just AI deployment. Take a holistic approach and assess your entire machine learning lifecycle. Think MLOps alongside questions of how can I deploy more quickly and effectively.

@victorthu I agree. Just because you have a AI hammer, doesn't mean every problem is an AI nail.

A4: Data drives change! Organizations need to look at how processes are actually occurring. This is done by looking at the timeline of events in a process. It can also find blind spots and common bottlenecks prime for AI fixes

A4: Invest in DL and true AI now as a differentiator, as ML know-how and execution will become more standard and buyable.

(edited)

@Adewunmi It's true that AI strategy has to consider the whole. It's not a stand alone project.

@BCorrIoT I always love this saying. People think they solved it all with one tool - they say it takes a village and it does, it truly does

@BCorrIoT Hard agree! And I think an excessive focus on PoCs can sometimes make it easier to miss non-viable ML projects/ use cases

@RyRaiker " Data drives change!" Basically the motto of our era.

And the other thing is, don't fall in love with super high accuracy with your models. Sometimes the actual business results between 80% accuracy and 90% accuracy is not material. It's better to just deploy them!

beyond all that .. maybe “keep it simple” - You know … “hey team let us not try to boil the ocean” even small AI wins can make a major impact on the enterprise

Yes, and that means data ingestion, it means recognising and planning for differences between the conditions in your development environment versus production. It also means taking a more cross-functional team approach to developing AI solutions

Exactly @RyRaiker!! You can definitely get AI wins starting with taking smaller bites.

A3: Getting overrun by data and not first solving enterprise-level data management. Assembling strategic talent in data science and analytics.

A3: 1) The over-romanticization of free open-source tools to deploy AI/ML. It’s hampering enterprises from getting any real ROI.
2) The scarcity of talents and the difficulties of hiring the right talents.

@victorthu Talent scarcity is HUGE in AI. Expensive talent.

A3: #AI pain points range from data reliability, AI sprawl, auditability & ethical challenges around privacy & potential algorithm biases. The first step is to ensure you have reliable, unbiased data while adopting proper #DataOps & #AIOps best practices.

A3: I think 2 big pain points here are (1) not having standardised practice for MLOps (2) not being able to take PoCs into production

Exactly @JamesMaguire !! And when you couple of that with people who want to build their own MLOps platform with the scare talents. It's a recipe for failure.

A3: Complexity is slowing progress along with wrapping your arms around all the data. The projects expand scope and encounter model drift and to some degree trying to get to a perfect model.

A3: I think another emerging pain point (which in some ways is an expansion of bullet 1) is not have clear frameworks for deciding when to build or buy off-the-shelf. Is your use case generalizable enough to make off-the-shelf viable? If not, can you make it so?

I agree, scarcity of talent all realms of data engineering and science is a major problem.

@BCorrIoT Totally agree on algorithm bias. It's a real challenge.

@Adewunmi And deciding to build is fraught with hazard. Big expense.

A3: I agree with issues others' have raised, I think in the past #DataOps has been underestimated. I think that's changing now. Or at least orgs are waking up to its importance.

Yes, I agree but understanding and clear articulation of business value is the starting point for deciding whether it's worth it. In our consulting practice, this is one of the questions we make sure to ask customers or help them work out.

A2: Companies’ comfort level with AI is still very early. I was recently talking to an enterprise that is known to have a massive AI team and have done a lot to incorporate AI in their business. Only to find out most of the AI models are still in the lab.

A2: #AI was once considered a back-office fundamental. But with the accelerated pace of digital transformation, #IT service & operations teams are using #AIOps to become more agile & proactive, to better anticipate challenges.

@victorthu Not surprising at all! But perhaps a waste of money??

A2. Companies are starting to get more comfortable with AI but it still has a ways to go. A recent Forrester study revealed that in 2019 only 54% of companies were using some form of AI, in 2020 that number grew to 69%. So again movement in a positive direction.

I think there's a spectrum here. I thinks orgs are definitely comfortable with the *idea* of AI. So there's maturity in that sense. I don't think we have hit maturity when it comes to identifying all the possible ML use cases within their operations.

A2: The #AI market is still very nascent, especially when operationalizing #IT.

@BCorrIoT #AIOps definitely on the upswing.

A2: Having said that, conversations about AI are getting easier, more realistic and more holistic as well

Yes, @JamesMaguire, there's definitely a sense of concern especially the lack of ROI with such a huge investment. So if a large company is struggling, the smaller ones will feel the pain even more acutely.

A2: They’re getting practiced in ML mostly. DL and true AI are frontier tech far from commercialization across the market.

@victorthu AI still feels like a tool for larger enterprise. But the SMBs are eager to catch up.

A2: By holistic, I mean customers are more open to thinking about AI as much more than primarily model development.

Yes @Adewunmi , very true. Compared to a few years ago where many AI vendors are selling vaporware vs. today more enterprises are better educated on the topic.

definitely. A recent study showed that 41% of enterprise IT use over 10 tools for performance management and monitoring. Hence the need to introduce #AIOps.

A2: Companies have become more comfortable using AI within the past 3-4 years, mainly attributed to the popularity of RPA. But, by itself RPA is not AI

definitely. A recent study showed that 41% of enterprise IT use over 10 tools for performance management and monitoring. Hence the need to introduce #AIOps.

@BCorrIoT That is a blizzard of tools. I think larger companies in particular want "best of breed" on everything. Which does drive AIOps.

@BCorrIoT Companies have also become comfortable using chatbots, but those who are digital natives spot them right away and the younger population gen-z and millennials tend to hate them! #ai #eweek

A2: maturity comes from collaboration and an ecosystem but not being locked to underperforming tools.

@RyRaiker The problem with chatbots is that they are artificial but not intelligent ;)

@RyRaiker not to mention increased collaboration between business and IT professionals in digital business initiatives demands new practices, policies and technologies.

A1: The over-proliferation of data, the inability of humans to leverage data at scale, and the great potential untapped data represents, including solving true human problems.

Q1 - Democratization of technology has made AI easier to consume. This has been demonstrated widely with the rise of #ConsumableApps #eweek

@Chris_Ehrlich Agree on over-proliferation of data -- for sure!

A1: AI for IT Operations (aka #AIOps) is a trend we are seeing heat up in the #AI market. AIOps applies AI and ML, & #bigdata analysis to automate #IT operations making IT teams more successful & efficient.

A1: The number one thing for AI is the need for transparency. By turning the AI black box into a glass box, individuals can turn insights into actions and make better decisions. This empowers humans and business to be more efficient and better serve customers

@Chris_Ehrlich data is everywhere, now machines can finally leverage it to do more while maintaining some of the same architecture and framework.

A1: We see that the market is moving more so into MLOps. What used to be known as model development, data gathering, monitoring, etc are beginning to adopt the same term.

@RikChomko But is it a glass box? I think it's still unclear for many.

I guess that's what I'm saying is that it needs to move towards being a glass box vs black box which is currently is now.

A1: Agree with @victorthu's point about the growth (or explosion even!) in MLOps

I agree, @JamesMaguire, it doesn't have to be a glass box. Just like most software today is not a glass box. It's a matter of understanding what your models are doing.

A1: Hyper automation and an explosion of data across IT and OT.

if it was glass .. maybe people wouldn’t be so afraid of it. It is these unknowns that keep people from adopting innovative AI technologies.

@RikChomko Yup, I'm with you -- positive moment is key.

A1: I think another (continuing) trend is the push for more data and more compute. We're really seeing this in the NLP space with the trend in LLMs showing no signs of stopping yet

@RyRaiker And I think some are uneasy with it -- particularly workers who feel threatened.

new research and development in AI shows that workers can be assisted rather than replaced - this information should be spread far in wide. Every human worker should wonder why they don’t use an AI assistant at work.