James Goulding hears how Arrow ECS is helping channel partners profit from AI
Global technology solutions provider Arrow Electronics held its inaugural technology conference at London Olympia on May 8. With exhibitors and speakers from across the AI technology ecosystem, Vision 2019 is an important part of Arrow’s strategy to become a leader in this emerging market. At the event, Technology Reseller met up with David Fearne, Arrow Electronics Global Practice Leader for AI and Data Intelligence, to find out more about how its distribution arm, Arrow Enterprise Computing Solutions (ECS), is helping channel partners profit from AI. We started by asking him why Arrow chose AI as the focus for its first Vision event.
David Fearne (DF): “We feel AI has come of age and moved out of the experimental phase. The tooling to do the data science – the stuff that lets you model and build these clever AI systems – has been growing in maturity for many years. What hasn’t been so mature is the set of technologies that lets a business take advantage of a piece of AI software, machine learning or whatever, and deploy it in a usable way. That has now come to maturity, so much so that putting it into enterprise adoption is no longer seen as the risk it was. In fact, it is now seen as disadvantageous to a business not to adopt AI in a big way.
“In that sense, the tables have turned. AI has come of age very quickly – it has done in 5 years what cloud did in 15 – and the balance between risk and reward is now even, if not tilting more to reward. Organisations are looking to partners to provide advice and guidance around AI, just as they have done for cloud, and networks before that. They want to know how they can adopt AI strategically, repeatably and with scale.
Technology Reseller (TR): It has to be done strategically because there are complex considerations to do with data privacy and ethics.
DF: That’s the thing. One of the biggest problems we have had with AI from Day One is that while we can build all these fancy models, it is difficult to explain how they come to their conclusions. The ability to put in a framework and tooling that, in the case of insurance scoring, would enable us to say ‘You didn’t get insurance because you didn’t hit these criteria and you scored low here’ has been a long time coming. The front end of AI, the machine learning piece, is relatively simple; it’s the eco-system of tools that go around it that is more difficult.
I liken it to virtualisation and cloud. On day one, virtualisation was a server and some virtualisation software that gave you VMs. That’s not a cloud. A cloud is where you have provisioning and monitoring and logging and clever networks that optimise themselves. At the moment, AI is just a box you can do some machine learning with. To make it into something that is appropriate for the enterprise, you need an eco-system of tools around it that enables you to do all the clever machine lifecycle management, automated testing and so on.
TR: And you provide your resellers with that eco-system?
DF: Our value in this space is expertise and understanding. I run our global data intelligence practice and my team’s role is to look at the latest developments in this space and how they translate into the channel. We are always on the look-out for net new vendors – not the big anchor vendors like NVIDIA or IBM, but the glue vendors that stick it all together, that take these disintermediated parts and glue them together into something that feels solid and production-ready.
There are very few end-to-end solutions in AI, which is why Distribution is so important. We have huge value in a complex and disintermediated market, because we have the ability to take this, this and this, bring them together, package them up and deliver an outcome. That’s never been more important than with AI.
TR: Are your resellers able to meet customers’ needs in relation to AI or do they lean on you?
DF: In these new technologies, they lean heavily on Arrow. One of our biggest roles is enablement. We are a scale organisation. We don’t work well if we have to hand-hold every single deal a reseller goes through. That’s not what resellers want either. The resellers that are successful and profitable are the ones we help to become self-sufficient, who use us for fulfilment or occasionally when things get complicated come to us and say ‘Can you help us with this one?’.
We work really well when we can scale and repeat; when what we are trying to deliver isn’t a snowflake. The problem with AI is that every project is a snowflake to a degree, because people want it to do very specific tasks, which are ever so slightly different to the way others do them. What we try to do is develop solutions that are repeatable and then work with service organisations to deliver the specialisation.
Currently, we are working on one with Microsoft that uses the Vision AI camera developed by our global component colleagues. Essentially, this lets us take the cameras, train modules and provide the whole lifecycle management for a vision solution, which could be used to track people, say, or monitor production lines. Because it’s built on the cloud, it can scale to thousands of cameras and gateways.
TR: Presumably, AI is still a step too far for many of your partners and their customers.
DF: Yes, and that’s why we have a whole strategy about how to prepare for AI. We call it the Arrow Enterprise Data Strategy and it provides a framework for big data that can help you break down data silos in organisation, which is necessary before you can adopt AI at any scale. If you just turn around and say ‘I‘ve got all this data in an Excel spreadsheet, in an Oracle database, in a Microsoft database’, you are going to spend more time trying to refine that data than delivering a valuable outcome.
Our framework for big data is really simple. It has four steps: ingest, transform, store and explore.
1 Customers start with ingest: what data do you need and how do you get it? If you work in sales and are trying to figure why customers aren’t buying, it is really useful to ingest marketing data
and social data as well. So, how do we ingest and break down those data silos in sales, marketing, HR, purchasing and manufacturing?
2 Next is transform: once you’ve got that data, how do you transform it into something more valuable i.e. information?
3 The third piece is store: how do you store that information compliantly, how do you make sure you are not infringing GDPR, how do you make sure that over time you destroy that data programmatically, so you don’t hold on to it beyond the scope of the different data projects you are running, how do you make sure that data is available everywhere you need it?
4 The last piece is explore: how do we make that data available to as many decision-makers as possible? Previously you might have had one person with an Excel spreadsheet running some formulas, but if you have ever picked up someone else’s spreadsheet and tried to understand what they were doing you will know how difficult that is. The portability of Excel spreadsheets is terrible. It would be really valuable if instead we provided you with some visualisations that could inform the widest possible cohort of people in their decision-making.
That is what we call the Arrow Framework for Data.
If a partner says ‘I am just not ready for AI’, we would go to them with our data strategy. We would say ‘All your customers want to talk about data. Here is how we can help you develop a story around data. You start the story here, then when you get here, you can say: ‘Now you’ve got your data in order; now you’ve got an industrial data pipeline that lets you start to do more intelligent things with data, we can start to talk about AI’.
TR: Is there a lot of demand from channel partners for help in developing stories to take to their customers?
DF: Absolutely, it’s probably the single biggest thing we get asked for, net incremental today. If I am asked what our partners ask for over and above what we have been doing for them for many years, it is absolutely around this.
Partners that think there’s still huge value in selling a product with very low-end services like rack and stack and initial installation are going to have difficulty going forward because the reality is end users are not as bothered about the brand of technology as before. Technology has become so ubiquitous that today it’s a case of ‘I need a solution that delivers this outcome and it needs to be under this SLA and it needs to be pay per month or scale up/scale down or fit this budget. Everything else I don’t care about’.
I always use the analogy that we need to be selling cars, not engines, and I think a lot of our partners are still trying to sell engines. I have opened a couple of events recently by asking ‘Who in this room drives?’. Everyone puts their hand up. I then ask: ‘Who knows the manufacturer of their brakes?’ Nearly everyone puts their hand down – being an IT crowd there are always a couple of geeky guys who do know. ‘Why don’t you know that, they save your life every time you get in the car?’ All of a sudden the cogs start to whir and people say ‘Actually that’s a good point. I don’t want to be a brake manufacturer as a partner. I want to be Ford or Toyota; I want to deliver the outcome. I want customers to say I want it red, I want it to be fast or economical or to carry lots of stuff’. It massively ups their value from being a brake manufacturer to selling the whole car.