It would be fair to characterize Ayasdi as a pioneer in enterprise AI. Our company is an early thought leader in this field, with proven success at many blue chip customers , supported strongly by major investors , felicitated by stellar industry groups and with many peer-reviewed publications detailing the incredible breakthroughs. Our brand of AI is distinctive (topological data analysis ) as is our application of it (intelligent applications ) and the enterprise class problems we focus on. Along the way, we have learned a thing or two about how to get started in AI, how to build momentum in AI and ultimately how to succeed with AI. We share these with customers and prospects, but wanted to be sure we did so more broadly. Question: Should I begin with the Innovation/R&D team to begin adoption of AI in my enterprise? On some level this seems logical - R&D teams are good a vetting new technologies and determining organizational fit. In reality though, R&D teams generally don’t solve real business problems. They often use synthetic data, and focus on laboratory style “tests” of software, rather than focusing on driving real business ROI. The result is that quite often, there is no path forward. Instead, it is far better to find the problem, then apply the technology. How do you find the appropriate problem? Look for challenges that:

1. involve complex data,
2. that is rapidly changing,
3. often with difficult to spot problems (weak signals),
4. where the operational process is labor intensive and slow and
5. where accuracy and operationalization are both critical. Answer: Tackle real problems in your lines of business that allow you to pilot AI, and provide you a vision for the operational next step. This has the added benefit of paying for other AI projects within the organization. We have seen customers obtain >10MM py benefit from singular use cases, which ends up funding other AI projects.

Question: When should I start - I don’t think I am ready yet. The time is now. Your competition is not waiting around. Be prepared to start, and then iterate quickly, committing to fast timelines. The default in larger enterprises is to look at projects in terms of quarters, if not years. Organizations can move faster particularly when faced with a crisis. As we have pointed out , waiting to see how AI develops is asking for a crisis. From a technology perspective the results come quickly - it is the data preparation that often takes the longest (along with standard enterprise on-board processes such as drug tests, fingerprinting and background checks). Putting great resources on these projects ensures that the organization sees value faster and learns faster. Answer: Learning is a ladder, attack the rungs. Question: I’d like to start, but my data is not clean yet. Garbage in, Garbage out, right? Should I wait for clean data? There really is no such thing as clean data, and there never will be. In general, there is a tendency to overthink the data preparation step. Granted there are best practices that need to be followed, but ultimately we find many enterprises are overly worried about bad data. With technologies such as TDA, there is still a lot of value even in the data you think is marginal at best. Nulls, sparsity, column heavy data are all elements that yield to our technology. As a result, you can actually extract a lot of value before you engineer your systems to account for the shortcomings and improve the quality (thus generating even more value). Answer: Start with what you have, your data will does not need to be perfect. There is already vast value in what you have. Question: What else should I be prepared for? The whole point of deploying AI is to observe process transformation and automation. As with all disruptive technology, be prepared organizationally for process change. Run these new systems in parallel with your current processes where they exist but be prepared to change your business workflow, and resource mix. In some cases, like our AML work , we have invented or so radically re-imagined the existing process that it doesn’t matter, but in most cases it will. A great example is our cyber work at DoCoMo where they kept their existing Splunk powered process and layered our process on top to catch the “unknown unknowns”. Ultimately, most of our applications require a change in process. For example, in our work with the old process built the risk model, then met with the business, then when back to the modeling. The new process consumed a fraction of the resources and required a fraction of the time. Adjusting all of the inputs and output to match this acceleration is critical if you're going to realize benefits from machine intelligence. Answer: Process change is implied with AI. Prepare your organization for it. Know that this is worth it - there is tremendous value in AI - and your peers are already benefiting from it. Question: But how do I systematically scale AI within my enterprise? The AI technology wave is upon us similar to how virtualization, cloud and big data have been. The time for casual experimentation of AI is over, AI is here to stay and must be treated as a first class citizen within the enterprise. Amazon, Google, Facebook, Microsoft have demonstrated this first hand. They have methodically buried their competitors over the last few years and increasingly look across the table at each other. Enterprises that are serious about AI (which should be everyone not looking to exit the business) need to be setting up a AI Center of Excellence (COE) to manage, optimize and scale their AI efforts. A COE is designed around scaling intelligent applications by providing governance, training, leadership and measurement. Answer: Set up an AI center of excellence where best practices are developed, process change is accelerated and prioritizations are made based on operational readiness, business need and other considerations. The next generation of leadership, and next source of business growth, will come from this center. Staff it accordingly. These steps are not a recipe for success, but they are a guide, a checklist to facilitate the transition from a linear world to an exponential one. While every organization should tailor their intelligence strategy to their distinct business needs, the blueprint offered here represents the knowledge accumulated in close to a decade of deploying intelligent applications for large, complex enterprises. In the coming years, every analytics company will claim to be an AI company. Such claims will make it difficult to distinguish truth from fiction. Our intelligence framework outlined in discover, predict, justify, act and learn presents a way to assess the validity of those claims where these guidance points offer a way to implement them. It would be fair to characterize Ayasdi as a pioneer in enterprise AI. Our company is an early thought leader in this field, with proven success at many blue chip customers , supported strongly by major investors , felicitated by stellar industry groups and with many peer-reviewed publications detailing the incredible breakthroughs. Our brand of AI is distinctive (topological data analysis ) as is our application of it (intelligent applications ) and the enterprise class problems we focus on. Along the way, we have learned a thing or two about how to get started in AI, how to build momentum in AI and ultimately how to succeed with AI. We share these with customers and prospects, but wanted to be sure we did so more broadly. Question: Should I begin with the Innovation/R&D team to begin adoption of AI in my enterprise? On some level this seems logical - R&D teams are good a vetting new technologies and determining organizational fit. In reality though, R&D teams generally don’t solve real business problems. They often use synthetic data, and focus on laboratory style “tests” of software, rather than focusing on driving real business ROI. The result is that quite often, there is no path forward. Instead, it is far better to find the problem, then apply the technology. How do you find the appropriate problem? Look for challenges that:

1. involve complex data,
2. that is rapidly changing,
3. often with difficult to spot problems (weak signals),
4. where the operational process is labor intensive and slow and
5. where accuracy and operationalization are both critical. Answer: Tackle real problems in your lines of business that allow you to pilot AI, and provide you a vision for the operational next step. This has the added benefit of paying for other AI projects within the organization. We have seen customers obtain >10MM py benefit from singular use cases, which ends up funding other AI projects.

Question: When should I start - I don’t think I am ready yet. The time is now. Your competition is not waiting around. Be prepared to start, and then iterate quickly, committing to fast timelines. The default in larger enterprises is to look at projects in terms of quarters, if not years. Organizations can move faster particularly when faced with a crisis. As we have pointed out , waiting to see how AI develops is asking for a crisis. From a technology perspective the results come quickly - it is the data preparation that often takes the longest (along with standard enterprise on-board processes such as drug tests, fingerprinting and background checks). Putting great resources on these projects ensures that the organization sees value faster and learns faster. Answer: Learning is a ladder, attack the rungs. Question: I’d like to start, but my data is not clean yet. Garbage in, Garbage out, right? Should I wait for clean data? There really is no such thing as clean data, and there never will be. In general, there is a tendency to overthink the data preparation step. Granted there are best practices that need to be followed, but ultimately we find many enterprises are overly worried about bad data. With technologies such as TDA, there is still a lot of value even in the data you think is marginal at best. Nulls, sparsity, column heavy data are all elements that yield to our technology. As a result, you can actually extract a lot of value before you engineer your systems to account for the shortcomings and improve the quality (thus generating even more value). Answer: Start with what you have, your data will does not need to be perfect. There is already vast value in what you have. Question: What else should I be prepared for? The whole point of deploying AI is to observe process transformation and automation. As with all disruptive technology, be prepared organizationally for process change. Run these new systems in parallel with your current processes where they exist but be prepared to change your business workflow, and resource mix. In some cases, like our AML work , we have invented or so radically re-imagined the existing process that it doesn’t matter, but in most cases it will. A great example is our cyber work at DoCoMo where they kept their existing Splunk powered process and layered our process on top to catch the “unknown unknowns”. Ultimately, most of our applications require a change in process. For example, in our work with the old process built the risk model, then met with the business, then when back to the modeling. The new process consumed a fraction of the resources and required a fraction of the time. Adjusting all of the inputs and output to match this acceleration is critical if you're going to realize benefits from machine intelligence. Answer: Process change is implied with AI. Prepare your organization for it. Know that this is worth it - there is tremendous value in AI - and your peers are already benefiting from it. Question: But how do I systematically scale AI within my enterprise? The AI technology wave is upon us similar to how virtualization, cloud and big data have been. The time for casual experimentation of AI is over, AI is here to stay and must be treated as a first class citizen within the enterprise. Amazon, Google, Facebook, Microsoft have demonstrated this first hand. They have methodically buried their competitors over the last few years and increasingly look across the table at each other. Enterprises that are serious about AI (which should be everyone not looking to exit the business) need to be setting up a AI Center of Excellence (COE) to manage, optimize and scale their AI efforts. A COE is designed around scaling intelligent applications by providing governance, training, leadership and measurement. Answer: Set up an AI center of excellence where best practices are developed, process change is accelerated and prioritizations are made based on operational readiness, business need and other considerations. The next generation of leadership, and next source of business growth, will come from this center. Staff it accordingly. These steps are not a recipe for success, but they are a guide, a checklist to facilitate the transition from a linear world to an exponential one. While every organization should tailor their intelligence strategy to their distinct business needs, the blueprint offered here represents the knowledge accumulated in close to a decade of deploying intelligent applications for large, complex enterprises. In the coming years, every analytics company will claim to be an AI company. Such claims will make it difficult to distinguish truth from fiction. Our intelligence framework outlined in discover, predict, justify, act and learn presents a way to assess the validity of those claims where these guidance points offer a way to implement them.