FORGET about personalization, portals and push. Today's mantra in information technology is ''services.''

Most financial analysts have been skeptical about the proposed merger of Hewlett-Packard and Compaq Computer, announced last week. But those who see some hope have based their assessment on the prospect of the merged company's being able to build a business in providing information technology services to its customers.

The I.T. service business ranges from day-to-day system administration, to the more difficult, and more risky, challenge of system integration: making new technology work seamlessly with older ''legacy'' systems.

The more information technology a company buys, the more it has to spend to support it and use it effectively. Erik Brynjolfsson, an M.I.T. economist, estimates that for every dollar spent on information technology hardware, roughly $10 must be spent on additional expenses like worker training, systems administration, business process re-engineering and other services. It is not surprising that a lot of technology companies are hungry for a piece of the services market.

Hewlett-Packard, looking at I.B.M.'s success in this area, has been trying to expand its role in services for some time. It tried to acquire the consulting company PriceWaterhouseCoopers last year; even though that acquisition collapsed, Hewlett-Packard has continued to make deals with PriceWaterhouse and another big information technology consultant, Accenture. As for Compaq, last year its services business generated about 23 percent of total revenue -- a bright light in an otherwise dismal earnings report.

But is the service strategy the way to go? If the real estate market were so depressed that no one wanted to erect new buildings, would it make sense to say, ''Let's get out of construction and go where the real money is: architecture''?

But perhaps that isn't the best comparison. Maybe the better analogy is this: if the construction business is depressed, remodeling could still be a very good business.

The information technology industry had three back-to-back investment spurs during the latter part of the 1990's: telecommunications deregulation in 1996; fears in 1998 and 1999 of the Year 2000 computer problem, and the Internet boom in 1999 and 2000. Each of these events led corporate customers to stock up on information technology. Now, large companies generally have a modern, efficient I.T. infrastructure. The problem is, they haven't figured out how to use it very well.

Paul David, the Stanford economic historian, described a comparable situation in his classic 1990 article ''The Computer and the Dynamo.'' Before electricity, manufacturing plants were built around a shaft running down the center of the factory. All the power tools were run by belts connected to the central shaft.

When electric motors came along, they provided a more reliable power source, but initially had little impact on the actual production process.

It wasn't until Henry Ford and company started experimenting with assembly lines in 1913 that electric motors really became useful. Ford and his executive team were on the factory floor every day, tinkering with the line: speeding it up, slowing it down, raising it, lowering it, moving the tools around and rearranging production flow.

Just as the motor allowed management to tinker with, and improve, the flow of material across the factory floor, so does the networked computer enable managers to improve the flow of information through the organization. The challenge now is to learn how to do this effectively -- to engage in the kind of tinkering and experimentation that Ford did back in 1913, but now with information rather than automobile parts.

Ford was happy to show off his assembly line, and the idea spread rapidly. These days the agents of technology diffusion are consultants. Like everyone else, consultants are pretty bad the first few times they install a new information system, but once they figure out how to do it right, they make a bundle of money telling other folks how to avoid their mistakes.

The message hasn't been lost on manufacturers. Cisco, I.B.M., EMC, Oracle, Sun and other leading technology vendors are intensely interested in how their equipment is being used. If they see good ideas being put to use by their customers, they are quick to trumpet them as new, successful applications.

In information systems these days, there is no one big killer application, like spreadsheets or desktop publishing. The killer app has been replaced by hundreds of small applications that, cumulatively, lead to more efficient production.

A few months ago, I toured a major media company. In this organization hundreds of creative people worked on each project, and it used to be difficult to coordinate their efforts. Now the creative teams use a central database to keep track of projects. Notes from each meeting, each sketch, each sound bite and each color choice are digitized and sent to a multimedia database to which all team members have access. The database becomes a group memory for the team, making it much easier to coordinate production. As one team member said: ''Sure, it's critical. If the database goes down, no one can get any work done.''

Mass market retailing is another area ripe for improvement. As more and more transactions are mediated by computers, sellers have been able to capture vast amounts of information about their customers. Supermarket loyalty programs allow companies to quantify reactions to advertising campaigns, fine-tune in-store promotions and provide personalized discounts.

But in most industries, data collection has raced ahead of data analysis. The data are there, but they have to be sifted through and understood before they can be used effectively. Companies won't be buying much more hardware or software until they learn to make better use of what they have. So helping companies make the necessary adjustments might be a good business for the next few years. To repeat the earlier analogy: when construction is slow, remodeling may still be a good business.

But information technology services is a tricky business. It requires having energetic people out in the field, watching for good ideas, and gaining experience in putting them to use. Incumbents, like I.B.M. and Accenture, will have a big advantage over new entrants, like Hewlett-Compaq.

Second, customers are going to be wary of conflicts of interest. Customer concerns about I.B.M. Global Services' promoting I.B.M.'s own products have forced it to withdraw from the software application business. Companies with a strong hardware presence are well placed to provide maintenance services, but will find it difficult to compete in the more lucrative systems-integration marketplace.

Third, it's not easy to build an organization that provides hardware, software and services all under one roof, because they are three very different kinds of business. True, I.B.M. has done it, but the company had a long history of providing one-stop solutions.

Services are probably a good business to be in. But that is like saying being a chef is a good business -- as long as you know how to cook.