Patents are a measure of research output and IBM research has the largest number of patents as compared with any other corporate laboratory in the world. Senior Vice President of research Paul Horn currently heads eight of the IBM research labs across the world, including the one at IIT Delhi in India. Horn is credited with having come up with a model to link research and product development where the product teams fund research. Now IBM research is in the throes of another change process and Horn spoke to Vidya Viswanathan of BT about it. Excerpts from an exclusive interview:

What is the current change that is happening in the research labs now? Research and services innovation...

Let me put it in a particular context. The context will help explain the change. If you went back 25 years ago, and you asked what were people doing in universities in computer science, they were mostly working on hardware. There was not really innovative work going on in software. Some engineers were working on it, but most thought it was something to make hardware work. If you went to the IBM academy, they were all doing hardware. Today, yes, we are doing software. Even IBM academy understands that there is real intellectual depth in software. If you go to an academic institution, they are working on distributed information infrastructure. But if you go to the academy and see how many people are working on services, it is almost zero.

They will give you the same arguments about services as they did about software. There is no intellectual depth. This is about outsourced data centres. And there is also no academic discipline. However, in the next five years, services science will be a new academic discipline. You go to a computer science department and ask people what they are doing and very likely they are working with biology or doing pattern matching in bio-informatics. Or they are working with the business school and doing electronic marketplaces or pricing analysis. They are working on hard business and societal problems. I would call that services science.

If you as IBM research lab were to sponsor university research in services science, what would the projects be?

In May we are getting educators from across the world to a summit to help define services science and then help stimulate it. An interesting piece of history. In the '40s, there was no discipline called computer science in the universities. A couple of IBM researchers and professors in Columbia had a joint work in computer science, one of the first in the world. In those days, the Watson Research Centre was in the Columbia campus. Then the joint effort became a course and then syllabus. So, I think services science is going to be a huge emerging and exciting area.

If a telecom company comes and tells your researchers, solve this problem, your researchers would do it. That is what you are doing with Charles Schwab, right?

Our researchers are working with customers and innovating in the marketplace. What this allows us to do is to bring back very hard problems. These require deep intellectual depth. Let me give you an example. If you look at the Harvard Business Review about six months ago, there was an article with a provocative title IT Doesn't Matter by Nicholas Carr. If you think about it in a traditional way, you could argue that it matters less now than it used to earlier. But if you think that it is becoming the language of Biology or it is becoming the language of models of how you do business, it will matter more now than it ever did. It is solving harder problems than it could ever solve before.

"To be successful, it takes invention and innovation"

Earlier computer scientists couldn't tell you too much about manufacturing. But we know today that manufacturing can now be modelled to create new efficiencies. But think about taking it to a whole new level. You can take a company, componentise the pieces of that company's processes, then pick one piece at a time, model and optimise them so that you create huge value. You can take a company and re-engineer it in a simple way.

If I come to you and say that I can make your company 20 per cent more efficient and increase your profit by as much, you will pay. That is not a commodity. The key thing that is different from the past is that re-engineering a company was a huge undertaking. People were reticent because they had to go through a massive re-engineering and they did not get value. If you break processes into itty-bitty chunks that are very well defined, you can get return on investment in a very short time.

How is your PwC acquisition helping you in this?

PwC has become the Business Consulting Services division and they are the channel for much of the research work. They have connection with key level executives and understand business problems. So I have to get my researchers to understand business and get their business people to understand modelling tools. Three thousand of my people and a whole lot more on the other side. One of the tricks will be to scale the work we have in research. Ultimately what you want to do is pick areas-we call them micro-practices-where research is the unique differentiator and over time build expertise not just in research, but also in our consulting practices and have it become the services practice of the future. Today we are very good at creating products that are IBM's future because we have arranged a marriage between research and development. I cannot say that we have built IBM's services practice yet. The goal is not to have a few researchers consulting, but bring back problems and build new services business for IBM Global services.

So in future the Infosyses, Wipros and KPMGs are going to have to fight this model...

You bet. If I am correct, we can use the world's best technical team to differentiate our services. No-body can compete with us. We are already closing business because of this. In the first year from a dead start we closed with $100 million of business. And this year it is going to be a lot more. About 10 per cent of our research organisation work on services now.

Are you going to work on all industries?

No, the problems are not verticals at this stage, but key technologies used in many industries. One of the micro practices is stichastic optimisation with noise (read: uncertainty). That is used in transportation, optimisation of airlines, fleets...One of the early customers we had was Boston Coach. This is a large limousine services company. We had to optimise the number of cars and dispatchers, but also get the car at the right place at the right time. But the problem is noise. You get bad weather, delayed planes. Optimisation in the presence of that kind of noise is stichastic optimisation. That is finding efficiencies where there are a lot of uncertainties.

Does research have a revenue target?

The profits from a consulting engagement go to the consulting business. We only use a part to keep the research going. We will use the same financial system that we used to get the product development team to work with research. If the development team funded one dollar of research for a problem, I took one dollar out of my corporate funding. So they get two dollars worth of work. You have to remember here that if all we think about is utilisation then why will they come to research? Utilisation is more bodies even though more skilled. They will use their own people. So we focus on margins. If they use researchers we give them higher profits.

How is corporate research going to be managed now? Xerox is a model that everyone admired. Microsoft is yet to prove itself. But they have a different view from yours. They are hiring the best researchers in different areas.

That is absolutely right. In my view, for whatever it is worth, they are building wrong models. That is the model we had, the model Bell Labs had and the one that Xerox Parc had. To be successful, it takes invention and innovation. The old model is great at invention and lousy at innovation. If you have not built the channels, if you are not working closely with the customer, you have to build a development organisation that will develop your invention. It is time consuming. In our industry if you are not bringing products to the marketplace rapidly, they devalue. I believe that the ideal model for corporate research is balanced. It has long-term exploratory work and short-time work that build the channel to the marketplace.

What prompted this? Lou Gerstner or the downturn?

We have always done this, but we were not as good as we are today.

But you too did miss the PC revolution...

We missed a lot of things. We built the first router even before Cisco existed. I couldn't begin to enumerate what we missed. We invented the RISC architecture. But we were slow in commercialising it. You could ask why we missed this or that...Have you heard of IBM printers?

Did you personally miss anything?

As much as anyone else in the IBM company. However, now we are better than anyone. All other laboratories will tell you nice things that they have done in their products. The difference in IBM research is all our products. Not something to spill ink better or some new algorithm to check spellings. If you look at IBM, every product is built as a partnership between research and product development. Now my challenge is to do the same thing for services.

Nanotechnology is the next big thing, right? In what time frame?

You can use various self-organising material in semi-conductor (chip) manufacturing in existing plants today. But if you think of some new material that is going to replace silicon like the carbon nanotube, that is very far. So, there is a spectrum.

Where is the next bout of innovation going to come from? That will solve the current controversy of outsourcing to India, right?

You missed my main point. It is going to be in services! In business processes!