(FORTUNE Magazine) – The dilemma is so sharp, it almost hurts: Companies are desperate to cut costs as the economy slows--Wall Street will murder them if they don't--yet reducing Net-related spending could put them at a competitive disadvantage in the greatest technological revolution of our lives. It gets worse. Most CEOs have seen plenty of disasters--not just failed dot-coms but also traditional companies that have poured hundreds of millions into large-scale IT investments that came in over budget, late, and ineffective. Early results of many Net projects are at least as bad: They blew up or just didn't work. Few managers can call on experience or intuition to judge these investments. At the same time, there's a nagging fear that standing still will leave them further behind. General Electric says a weak economy is exactly the time to increase tech spending to widen a company's lead over competitors.

Prof. Larry Selden, of Columbia University's Graduate School of Business, has been grappling with this issue. In a new course, Internet Financial Management, he focused on how companies can build shareholder value through Net technology. In addition to hearing from tech experts, his students pored over real-world projects and problems offered up by class sponsors Applied Materials, Barnes&Noble.com, Travelocity, and Trilogy. We asked three of the lecturers how companies could improve the odds of getting returns from their Net investments in a slowing economy. Done wrong, such investments succeed maybe 10% of the time; done right, the odds rise above 50%.

--Geoffrey Colvin

GREG PAPADOPOULOS Chief technology officer, Sun Microsystems: Don't invent everything from scratch.

Look at E*Trade as an example. When you get on the Web and go to E*Trade, your request launches a little piece of computation that goes back into a tier of servers. These are application servers, which keep track of a user and implement the business logic. Now imagine an online store: If someone clicks on buy, what should the server do? Should it clear his credit card, check a database, generate a pick list--what should it do? The business logic on the application server holds the answer.

The application servers then connect back to existing systems, such as databases, or maybe an enterprise resource planning system, or a connection out to Visa to clear a credit card. If you look at how E*Trade leaped ahead early on, behind the application servers was an existing VMS system, Digital Equipment stuff. They have connected it to a Web-based front end.

You know what? That's exactly what you ought to be looking for. That way it's not inventing everything from scratch. Starting from scratch is a disaster. I've seen so many companies go down in flames doing this kind of thing. The really excellent stuff is where you're able to look at existing systems and processes and put in new business logic to get them to do the things they ought to do. The Web front end is really just a way of getting to that and expressing it.

So the value is not that you create a Web front end. The value is your intelligence in thinking about the business logic--and the logic implemented on the application server is hooking together a whole bunch of existing systems in the back end. The magic is all in the business-process innovation. Don't get distracted by what goes on in pretty Web pages.

TERRY JONES CEO, Travelocity.com: Marry IT and marketing.

Running a successful Internet site is a really tight combination of data processing and marketing, two departments that traditionally do not play well together. That's why more and more CIOs are coming out of marketing and vice versa, because that's what this is, data processing and marketing going on at the same time. It's important to focus on the end result and hold people responsible for making it very fast.

We're trying to get more two-man teams, a marketer and a developer, and they bolt themselves together and really move fast. We did this because I was starting to see Travelocity become like other large development organizations, a place where a committee of people write a big set of specifications that are so detailed they're practically programming language, and then give them to a programmer who codes them and produces a result that is different from what was expected, three months later. So we're trying to say, "Let's just have a couple of people go do this, because I don't know if it's going to work or not. Let's find out quickly and expand it over time."

JOE LIEMANDT CEO, Trilogy: Business process improvements drive Net investments, not the reverse.

Business processes are hard to change when they involve lots of potential pain and risk. So from a software and technology point of view, the entire game is finding software that lets you make hard business-process changes easier.

Take the automotive industry, where we've been helping Ford. Tons of reports say carmakers would save $2,500 a car if they moved to a build-to-order system. People should buy over the Web. Why can't they? That's a hard business-process change. There are legal issues, dealer issues, logistical issues.

But we can attack them one at a time. First issue: Before the Web, auto manufacturers didn't even know who their customers were. So the first step was to build the relationship with customers through a Website. You couldn't buy a car, but you could at least configure it and price it--except that you couldn't price it because the price is actually negotiated at the dealer. So then the next phase in the software was to allow a consumer to get a real price quote, based on actual transactions in their regional market. Now, if you go to FordDirect.com, you can get the price and complete the Web transaction.

There are 15 steps in this hard business-process change that can be addressed by technology. Addressing them will turn this from an infrastructure and technology decision to a business decision: Do I want to sell cars over the Web?

A major problem with many IT projects, unlike Ford's, is the big-bang approach most companies take. The project is set up so the company has to implement the software completely, going from hundreds of legacy systems to one or a few--taking years, costing hundreds of millions of dollars--before the first dollar of value gets returned. That's why CFOs hate these projects. But there is another way. We're doing a system with a major direct retailer that's intended to improve customer service, increase order frequency, and increase average order size. We're taking 5% of the call center to show that when they use this system, it produces those results. If we can show that, the CFO all of a sudden says, "This is now a profitable project, and I want to fund it." For most of these projects, the risk factor is not the technology. It's business alignment.

The bottom line is simple: Keep your eye on shareholder value and hold your wallet. Net investments can be powerful but are not magical. To have a chance, they must be aligned with your competencies, your customers, and your business.

FEEDBACK: gcolvin@fortunemail.com