The failure of computer technology to deliver on ambitious promises of bottom-line value, together with a number of conspicuous systems disasters, has led many top managers to question the notion that IT can be used to gain competitive advantage. They are not alone. A recent MIT study found little evidence for any direct link between investment in IT and improved bottom-line performance.¹ Thus, with something close to a sigh of relief, many senior executives have returned to the familiar, low-tech world of quality management, front-line empowerment, and even business process reengineering to drive necessary organizational change.

Reconfiguring skills and organization around the intelligence inherent in the technology works—and the benefits are huge

Their timing is unfortunate. In the past, IT’s troubling lack of business value has stemmed as much from a failure to reconfigure skills and organization around the intelligence inherent in the technology as from hands-on problems in implementation. New applications, however, now make such reconfiguration feasible. It works—and the benefits are huge.

Rebalancing human and machine intelligence can, for example, more than double clerical productivity, as well as liberate front-line employees to serve customers, rather than perform administrative tasks. And it can cut employee learning cycles from months to weeks and even sometimes days, enabling the rapid reshaping of established businesses. Indeed, some recently introduced direct financial services have actually been set up from scratch in nine months or less. Such capabilities offer significant first-mover advantages. But they also make traditional markets vulnerable to unexpected competition from new, greenfield entrants. As opportunity and as threat, they deserve a closer look.

The skill imbalance

In many companies, even routine operational processes have become too unwieldy for the human skills that struggle daily to cope with them. In one insurance company, for example, internal processes have become so complex that training cycles now extend beyond 12 months, and a large, expensive supervisory structure is needed to check that every transaction is correct. An important source of such excessive complexity has been the inability of many companies to replace their "legacy" systems—the old software and hardware portfolios built laboriously over the past three decades—with new and more flexible systems. For years, the seeming intractability of this legacy problem has pushed the cost, timing, and inflexibility of new IT investments to unacceptable levels.

As recent developments have shown, however, there are ways to tackle this problem by effectively leaving the legacy systems alone, bolting on new application packages, and then linking the two sets of software with interconnect technology.² But removing this kind of bottleneck has brought another—perhaps even more harmful—by-product of past complexity to light: a skill overload that breeds disenchanted front-line workers and cumbersome management structures.

When process complexity escalates rapidly, the demands on the skills of workers escalate more rapidly still

When process complexity escalates rapidly, the demands on the skills of workers escalate more rapidly still. This usually provokes a counter-

reaction—a visceral reluctance to move away from the "devil we know" in organizational form or operational procedure. And this reluctance, in turn, stiffens the resistance of already nervous top managers to the idea of reconfiguring their businesses in innovative—and value-creating—ways. This vicious cycle of reactions is perfectly understandable. But it is unnecessary.

Today, managers can leverage machine intelligence in a manner that allows a fundamental rebalancing of the demands on human and machine skills. This is because so-called "intelligent" software makes it possible to shorten the learning curves associated with new business processes. In the insurance company mentioned above, for example, the redesigned systems had, by intent, a user interface that could be learned in four weeks or less. So condensed a learning curve boosted the redesign’s acceptability, speeded its roll-out, short-circuited resistance, and raised management’s aspirations about the company’s ability to reach new markets with new business systems.

The ability to buy in quickly "learnable" process innovations lowers the traditional barriers to entry

At the same time, however, the ability to buy in, rather than laboriously build, quickly "learnable" process innovations lowers the traditional barriers to entry in many complex industries. Using these new technologies, greenfield entrants can often develop superior value propositions and build them into low-cost business systems in timescales and at investment cost levels previously undreamed of.

One European bank, for example, which created a new service aimed at a neglected customer segment in only a year, achieved profitability and market share within eighteen months of starting up. In the United Kingdom, a single technically literate entrepreneur built from scratch a personal financial services company that now has over two million customers. And in the United States, the largest mortgage processor is a once minor player that has rapidly overtaken all its competitors thanks to a dedicated focus on process innovation.

In each case, the competitive breakthrough derived from a rebalancing of human and machine skills at the front line. In each case, this rebalancing enabled the development of flatter, less hierarchical organizational structures, freed of the traditional requirement to exercise painstakingly minute supervision and control. And in each case, implementation was both fast and flexible. Toe-in-the-water incrementalism was nowhere to be seen. Nor was it missed.

Man versus machine

In many companies today, skill overload is the real bottleneck to competitively significant process innovation. "Brick wall" might be a more appropriate metaphor. One bank extended its service range to more than 60 products to take advantage of new opportunities stemming from deregulation. Previous experience with training, however, suggested that most branch employees could master one or two at most. That meant a wild proliferation of IT-based administrative support processes, little cumulative expertise in each, and—not surprisingly—major problems in service, quality, and front-line morale. Legislators, fearing that this complexity could be manipulated to the disadvantage of consumers, fuelled the problem by imposing new compliance disciplines. The predictable result: a further downward spiral in service, quality, and morale. Force-feeding employees with yet more technology-based process redesign only made things worse. People had neither the time nor the appetite to learn how to use the new systems effectively.

"Managing around"

Perhaps the most troubling aspect of skill overload is the extent to which its negative effects are mutually reinforcing. Consider, for example, the work of traditional claims processing in insurance—a highly complex activity, given the intricacy and uniqueness of each individual underwriting decision. It is also highly correspondence- and paper-intensive. The tedium of the job, a Herculean task of managing an inexhaustible in-tray, has usually caused staffing to be limited to middle-ranking clerical operatives on fixed salaries—and management style, to heavy-handed supervision designed to resolve the ceaseless flow of questions on minute points of detail. Over time, both supervision and tedium have made the process inflexible.

When new business needs arise, the tendency is not to redesign the process, but to "manage around" it

As a result, when new business needs arise, the tendency is not to redesign the claims process, but to "manage around" it. If improved customer service in claims management becomes important as a source of differentiation, why try to tackle the unresponsiveness of the process? Why not simply bolt on new layers of infrastructure instead—quick-response telephone units, say, or field sales support offices? If product proliferation has made underwriting more complex, why not bolt on third-party assessors? Indeed, for years, the counsel of wisdom has been to let the underlying process slumber on untouched. So what, if clerks still amend policies by marking up computer print-outs in red ink, leaving it for others to puzzle out what these hieroglyphics might mean? Let sleeping dogs lie.

This is, obviously, inefficient. But it is also habit-forming. And as with any form of addition, the more you decide to "manage around," the more extensive and elaborate—and cumulative—the necessary evasions become. It does not have to be this way. Rebalancing the mix of human and machine skills—that is, making effective use of the intelligence underlying technology-enabled process redesign—can break the habit.

One insurer, for example, has built a sophisticated new working environment in which all paper handling has been replaced by image processing. This has freed up nearly one-third of the time once spent on handling voluminous files. More important, the new system controls all scheduling and prioritization with sophisticated workflow management technologies that liberate clerks from having to manage extremely complex transaction streams. At the same time, careful attention to the micro-level detail of the operator’s job has reduced training periods to days rather than months. In addition, expert systems help categorize and resolve many claims decisions; where live experts are needed, their services are available on-line.

The step-change in efficiency of this claims system, which allows even archival records to be available on screen in less than 30 seconds, has proven extraordinarily attractive to the company’s brokers and customers. The system is accessible through field support branches and—by phone—through a fast-response center. What makes it attractive, however, is not just its technical capability, but also the organizational—and behavioral—changes it both enables and encourages.

Rebalancing skills can open up entirely new ways to run a business

But rebalancing skills between human and machine does more than promote efficiency. It can open up entirely new ways to run a business. It enabled a US-based provider of intelligent technology to the financial services industry, for example, to reconfigure its claim system for automobile insurance around a radically new customer service proposition.

Questioning existing working practices at all stages of the claims process right from the scene of an accident, managers built a new system around pen-based computers, with large data storage and expert-system-driven logic, so that assessors could give an on-the-spot estimate of insurance damage for any model of car on the road. This estimate would then be transmitted by radio link, using a digital picture of the accident if necessary, to underwriters for instantaneous approval. Finally, using sophisticated interconnect technology, it would be circulated throughout the company so that all legacy systems could be updated and the transaction completed.

This new system supports a lower-cost, as well as more disaggregated, business system and a shorter training period

In practice, this new system supports a lower-cost, as well as more disaggregated, business system and a shorter training period (less than 24 hours) for learning the necessary estimating tools. In addition, it eliminates the need for a career assessor force and makes possible the outsourcing of processes—maintenance of the vehicle damage database, for example—that can be undertaken more economically by third parties. But most important, it no longer forces customers—who, after all, paid their premiums up front—to submit complex claim forms and then wait endlessly for their money. The competitively powerful goal of this new system: a "pay at the pump" service that hands accident victims their checks on the spot.

The sources of innovation

No one doubts the competitive potential of such breakthroughs. Rather, the practical managerial issue has always been the agonizing slowness of the design-and-build cycles encountered in most de novo development projects. This conspicuous lack of speed is not just the result of poor project management. It has traditionally been inherent in the very analytical disciplines used to structure these projects in the first place.

Why should managers pin their hopes on the availability of impromptu creativity whenever they need it?

Systems analysis, the methodology on which most software engineering -as well as process redesign—is based, depends on the rigorous documentation and flowcharting of existing processes, followed by the brainstorming of possible opportunities to develop improved, computerized alternatives. Why, though, should managers pin their hopes on the availability of impromptu creativity whenever they need it? And why should they deliberately doom themselves to the slow, mind-numbing slog of developing a mass of complex process flow diagrams? (Systems "specifications" of this type can easily run to more than a thousand pages and take well over a year to prepare.) Bogging the design process down in so much detail, even if it is accurate, virtually guarantees that no one will recognize a radically new approach even if one comes into view.

When technology is a major enabler of new processes, real innovation tends to emerge during user application, not in the development lab

In The Sources of Innovation, Eric Von Hippel draws on ample case experience to argue that significant process innovation usually happens in a different fashion.³ Across industries and technologies, the evidence is consistent and clear: when technology is a major enabler of new processes—as, say, with scientific instruments—real innovation tends to emerge during user application, not in the development lab. According to T. Capers Jones, an international expert on software development, this conclusion is directly applicable to information technology. In big systems development projects, more than 40 percent of the eventual functionality is built through amendments and adjustments after a system goes live.

Armies of generalists do not usually deliver the most valuable kinds of process innovation. Niche solution providers do

Most technology-enabled process innovation occurs in the "living laboratory" of modeling, prototyping, and real-world application, where developers and users work in partnership. Sometimes this happens internally. Increasingly, however, it takes place in the growing legion of small, specialized "intelligent" software providers—expert firms with somewhere between 50 and 250 staff, a strictly limited product line, and an international client base. Armies of generalists, equipped with laborious and complex methodologies, do not usually deliver the most valuable kinds of process innovation. Niche solution providers do—and at a far lower cost. For users, the key problem is the narrow application focus of each provider. If these applications are to serve as the building-blocks of redesigned, value-adding processes, users must be able to connect them together quickly and with minimal additional investment. Fortunately, the growing acceptance of "open systems" standards has brought a new group of vendors into the marketplace, which not only provide effective interconnect products and services, but also work with many different niche software suppliers to ensure that their products can be easily linked with one another in user-built networks.

Off-the-shelf access to most of the key ingredients needed to short-circuit the work of redesign lowers an industry’s entry barriers dramatically. It also challenges established notions of sustainable competitive advantage. When process innovation is equally available to both existing and new industry players, advantage depends on a company’s speed in commercializing it, not on creating—through massive investment—a unique process infrastructure.

Accelerating the commercialization of process innovation depends, in turn, on the rebalancing of human and machine skills. Such a rebalancing implies a very different approach to the lengthy planning cycles that have traditionally driven process improvements. Six months of strategy definition, followed by a year of process reengineering, and then two to three years of implementation, are still the norm in many large, established companies. Capturing first-mover advantages, however, means leveraging the "art of the possible" to collapse such extended timeframes—that is, linking process redesign to different combinations of lego-block business applications in ways that slash development and implementation cycles to a year or less, even for complex processes.

Building flatter organizations

Intelligent technology makes far fewer demands on middle management for supervision and operator support

As noted above, intelligent technology-enabled processes make far fewer demands on middle management for supervision and operator support. To oversee its intricate manual workflows, one traditional financial services company had a supervisory structure comprising at least four layers between business unit management and front-line staff. Though unwieldy and expensive, this infrastructure was essential—given the training required to achieve proficiency in so tortuous a set of administrative processes. In fact, offering supervisory management posts to skilled employees proved the only way the company could retain the people it needed to monitor and authorize its day-by-day transaction flow.

Intelligent technology makes this kind of compromise unnecessary. A workflow management system can monitor and control every operator’s performance against agreed customer deadlines. On-line, on-screen, step-by-step guidance can dramatically reduce the error rate. And where supervisory assistance is still needed, requests can be routed to a small number of specialist managers who serve the entire organization, not just a single geographic unit.

How, then, should managers approach the task of redesigning processes around intelligent technology? The experience of companies well along in this effort suggests five rules of thumb:

Start with the hypothesis that absolutely no intermediary structures are needed

1. Adopt a "zero-based" approach to management infrastructure. Given the diminished need for supervision, start with the hypothesis that absolutely no intermediary structures are needed, and then put in place only those required to make the new process(es) perform to spec. Certainly, such an approach is better than gradually trimming away an excessively large infrastructure that will fight tooth and claw to justify its existence. Turning conventional assumptions on their head can pay huge dividends. In one new direct bank, over 95 percent of the staff are front-line service operatives. In a newly established but highly successful insurance company, there are only nine managers above supervisory level in a total workforce of a thousand people.

2. Understand where to put the front line. In many industries, from servicing elevators to providing insurance by phone, there is a strong movement toward delivering services directly from a centralized location rather than through a traditional branch network. These centralized channels are much easier to control. Supervision ratios and costs are lower—and quality significantly higher. In fact, error rates fell by a factor of five in a newly established direct service unit in a retail bank. Moreover, the "factory" nature of centralized processes allows both for a smooth flow of work and for flexible, often part-time staffing arrangements, which together can improve efficiency by 50 percent or more. One financial institution estimates that it can manage five times as many cross-selling customer relationships with a centralized, as opposed to a branch-based, salesforce.

Over 80 percent of personal financial service users under 35 now prefer direct to branch-based operations

There is still a role, of course, for branch operations, especially where customers prefer face-to-face interactions. But thanks to the service and quality improvements of direct operations, even these preferences are shifting, especially among young consumers. A recent US survey showed that over 80 percent of personal financial service users under 35 now prefer direct to branch-based operations. This means that it is no longer obvious precisely where to put a company’s front line. Intelligent technology has made it an eminently movable frontier.

3. Balance electronic and organizational linkages. So-called "horizontal" structures can help optimize the day-to-day linkages across cross-functional processes, but their organizational consequences are often difficult, if not impossible, to resolve by conventional means. One retail bank, for example, tried to coordinate customer response through its direct marketing, telesales, and branch-based channels around a single highly-focused customer service proposition. The result was exceptional: a near doubling of market share only weeks after launch. But subsequent attempts to institutionalize this new approach foundered on the lack of interdepartmental cooperation fostered by the groups’ different skills, management styles, and cultures. Recognizing this problem, a European bank is attacking the coordination issue with a mixture of technology- and organization-based linkage mechanisms. It is using centralized customer diaries and case management, driven by intelligent marketing systems, to coordinate the actions of disparate business units into a unified strategy geared to boosting sales effectiveness. At the same time, it has also coordinated these units at the geographic, micro-marketing level better to satisfy the specific needs of local customers. This combination of organizational and electronic linkages is essential when companies need to manage a wide mix of sophisticated products and services without sacrificing the economies of scale of a large customer base.

4. Manage the availability of expert knowledge. Experience shows that expert systems and artificial-intelligence-based techniques can embed real expertise in a business process. But there are technical and behavioral limits to the amount of intelligence that can be supplied in this way. Large systems with thousands of decision rules absorb vast quantities of processing power. More important, they quickly grow so complex that they become inflexible. One of the most sophisticated expert systems for financial planning failed because its users, unable to figure out how it made its decisions, refused to bet their businesses on a "black box." A "legacy" machine brain can be as—if not more—dangerous to a business than a legacy process.

The wider application of machine-based expertise will put a growing premium on the availability of human experts

In general, therefore, the wider application of machine-based expertise will put a growing premium on the availability of human experts. Traditional motor insurance providers, for example, know that there is a big difference in profitability between various categories of risk at the micro level, but few ever build the capability to carry out a meaningful segmentation on the basis of the real risk/reward ratio of individual customers.

Recently, a new entrant designed its operations around a highly sophisticated risk segmentation system, which could be updated in real time against changes in market pricing and then used to inform the minute-by-minute underwriting decisions of its front-line telephone operators. The user friendliness of this system enabled the company to operate without hordes of specialists or cost-heavy management structures. Instead, it relied on a small number of exceptionally skilled underwriters to adjust the expert system’s rule base in real time as market conditions changed.

5. Take a fresh look at outsourcing. This new-found ability to use IT both to link separate lego-block businesses and to make scarce expertise available from remote locations raises the question of which processes and skills a corporation must own—and which it can buy from external sources. As some observers have long predicted, the permanent cores of many organizations are withering away as work gets outsourced to electronically linked cottage industries and part-time specialists. In such a world, external low-cost providers can carry out the commodity transactions that are not central to a company’s core value proposition.

Management practice has been understandably slow to adjust to the possible effects of intelligent technology. This is, however, a dangerous state of affairs. Traditional planning horizons and multi-year implementation programs are seriously out of kilter with the lego-block businesses that can now be quickly assembled from the off-the-shelf offerings of niche solution providers. Moreover, traditional process design efforts seriously underestimate the options generated by—as well as the competitive threats that can suddenly emerge from—a root-and-branch rebalancing of human and machine skills.

For better or worse, the sustainability of competitive advantage has fundamentally changed: the kind of innovation cycle that makes each new generation of a technology-intensive product like a PC obsolete within 18 to 24 months already challenges—or will soon challenge—the business systems and strategies of whole corporations.

There is, of course, no way to know for certain how these challenges will play out in any particular case. In an environment shaped by intelligent technology, there are few, if any, hard-wired solutions. Managers will not discover what to do by deducing proper courses of action from first principles. Instead, they will need to experiment and prototype and pilot-test—proactively, aggressively, repeatedly. They will need to block off limited geographic or business areas—that is, create a kind of ring-fenced greenfield situation—and try things out. They will need to rebalance skills and assemble the building-blocks of their businesses in a variety of ways—and then see what works.

Being intelligent about intelligent technology does not mean knowing all the answers or doing all the analyses. It means being willing to roll up your sleeves and explore.