Human Resources 2.0: How To Do More With Less

Do-more-with-less - that's the new mantra for managing human talent in the economic downturn. Amidst layoffs and reduced venture capital financing, companies are seeking innovative solutions that balance their continued need to solve complex problems with their hiring and budget constraints. In this article, we will teach you how to deploy an efficient model for managing your company's staffing needs. Whether you're a garage start-up or a multinational corporation, the Distributed Agent Model allows you to stretch your budgets further and dynamically grow and shrink based on your business needs.

Managing Your Human Capital
While the Distributed Agent Model may sound like something from a spy novel, companies such as InnoCentive.com and DuPont are using it to tap into brainpower outside their company. Probably the most popular incarnation of this model is crowdsourcing, where a problem is broadcast to a large audience with the hope that someone in the crowd will solve it. Here is a step-by-step approach to using the Distributed Agent Model within your organization.

Step 1 - Define and Scope Each Task
It's useful to think of your company's complex problems in terms of tasks that can be completed in a sort of global factory with workers distributed around the world. For example, an online travel site might use remote agents to act on complex hotel reservation requests that a computer simply couldn't handle such as "make a hotel reservation that would be perfect for a Valentine's Day surprise." Clearly a computer couldn't define what "perfect" meant, but perhaps a distributed set of workers, organized effectively, could find a great chic boutique hotel that fits the bill. The important pieces of this example are that tasks can be as complex as your workers can handle but they should be self-contained, meaning everything needed to complete the task is tightly packaged up within it. What you're looking to avoid when designing your task is the kiss-of-death for a distributed agent model: manual intervention. AskSunday.com, a New York based virtual concierge service, does a great job of allowing remote agents handle complex tasks such as booking travel plans.

Step 2 - Source the Right Workers
Seek workers who possess the core competencies you need for the task. Unlike conventional staff hiring, since you are recruiting for a specific task, "hiring up" or opting for over-qualified candidates could result in an agent population that is unenthusiastic and less motivated.

Consider intrinsic rewards in addition to, or even instead of, financial compensation. Finding a way to give your agents a pat-on-the-back, even a "virtual" one, goes a long way. ThisNext.com, a social shopping site which uses human agents to source unique online deals, does a great job of this. ThisNext.com rewards its agents with both novel "thank-you" gifts as well as kudos in the form of public recognition of their successes for all members to see.

Step 3 - Automate the 99% Case
Most of what your agents do should be completely autonomous and not require any manual intervention. The moment your system requires manual intervention, you have introduced a scalability problem into the system. For example, for a task such as proof-reading an essay, if a staff member is required to skim through the finished product, you have just created an expensive bottleneck. Of course, occasionally you will need to manually intervene in order to take corrective action or to re-assign a task - make sure that this is truly the exceptional case. TutorJam.com, an online tutoring company, has developed a complex Java-based backend system that allows hundreds of tutoring sessions to take place simultaneously without intervention. The system uses a combination of educators and automation to detect unusual situations.

Step 4 - Incorporate a Feedback Loop
When you have hundreds of agents "running around" performing tasks, you will need a quick way of assessing the successful completion of a task. Perhaps, having two agents independently completing the same task and comparing results might make sense. In many cases, simply having a "feedback loop" can signal that a task was completed successfully. For example, if the consumer can implicitly or explicitly acknowledge or even rate the quality of a completed task, a feedback mechanism is established. Be creative in defining your feedback mechanism - in many cases, a simple check of the volume of web traffic visiting the finished product or a quick 5-star rating system similar to Amazon.com or EBay.com may suffice. In both cases, a rating mechanism allows you to quickly evaluate the effectiveness of a particular worker or task.

Step 5 - Jump Start Your Learning Curve
This is where your strengths as a Web 2.0 company really come into play - agile deployment and fast product cycles. Your first attempts at a framework for managing human agents will be imperfect. The reality is that handling distributed human agents and the resulting problems could leave you scratching your head. Frequent releases and iterative upgrades will help you continually improve and refine your system and set you on the path to achieve your business objectives.

Tell Us About Your Experience
Our experience launching TutorJam.com, a K-12 online tutoring company, has shown us how flexible and scalable the Distributed Agent model can be in meeting the diverse educational needs of our customers. Let's get a dialogue going; Share your tips and stories with other organizations looking to capitalize on the growing trend of distributed agents, and tell us how you do more with less.

About the Author

Nathan is currently Vice-President of TutorJam, the premier online tutoring company. He has over 7 years of engineering experience at technology firms including Microsoft, Slipstream Data, and Texas Instruments. He holds a Bachelor's in Math and Economics from the University of Waterloo, a Master's of Science from the University of Washington and is an MBA candidate at Duke University.