The real promise of AI: inventing new work processes, not just automating the old ones

The vision of an agentic workplace has invaded our collective consciousness.  A world in which AI agents outnumber, outthink and outaction their human co-workers seems more than simply plausible, it seems increasingly likely.

Initial attempts to realize this vision have focused on automating tasks that follow predictable procedures and employ well understood business rules.  More specifically, AI agents in the vanguard of the agentic revolution are being trained to execute tasks commonly performed by software developers, customer service  agents, sales development representatives, IT support technicians, SOC analysts, HR recruiters, legal assistants, etc. 

Is AI-enabled automation truly a source of long term competitive advantage?  Early AI adopters may have a transient opportunity to improve the profitability of their companies by reducing labor costs, but won’t their competitors ultimately be able to duplicate their success simply by purchasing well trained agents with proven capabilities from one or more agentic marketplaces?

A more profound and enduring form of competitive advantage can be achieved by leveraging agents to invent new work processes that were inconceivable in the past due to staffing limitations, time constraints or the fine grained complexity of actual business operations.  What if armies of AI agents were able to overcome such restrictions and used to implement one or more of the following processes?

Consumer Marketing: Fine-grained Customer Persona Agents

B2C firms typically employ customer personas to ensure that their marketing efforts address the needs and interests of prospective buyers.  Historical customers with similar buying behaviors are placed in affinity groups.  A great deal of analytical effort is invested in determining the common characteristics of the individuals within each group.  These characteristics define the group persona and are hugely helpful in identifying individuals with similar characteristics who should be buying a company’s products but currently aren’t. 

The effort involved in defining and maintaining customer personas typically limits the number of personas a firm can realistically manage to several dozen or a few hundred.  AI can change all of that.  Agents can be constructed to represent the shopping proclivities of thousands of affinity groups.  Instead of limiting the number of attributes associated with an affinity group to several hundred and updating them sporadically, thousands of attributes associated with individual groups can be maintained and updated dynamically as new data becomes available.  The features of future marketing campaigns could then be tested across thousands of personas to optimize the revenue return of marketing investments. 

The multimodal interfaces of AI agents would also enable Marketing Managers to converse directly with individual personas to test new ideas or explore future customer preferences.  Marketing managers could even stage debates among different agentic personas concerning new product features or customer service policies. 

Maximizing Profitability: SKU-level Product Marketing Managers

B2C firms offering their customers a wide selection of retail products typically manage their marketing efforts at a product category level.  Flagship products that generate large revenues or are closely affiliated with the firm’s brand reputation may be managed at an individual SKU level but in most cases multiple SKUs are grouped into a product category and assigned to a Product Category Manager (PCM).

PCMs not only manage promotional and placement efforts.  They’re also responsible for making or influencing discounting decisions.  The timing and targeting of discounts plays a critical role in determining profitability margins, particularly during peak shopping seasons in the fall (back to school) and winter (Christmas holidays).    

Individual AI agents can monitor sales activity at a SKU level and employ ancillary information about inventory levels, inventory location and logistics expenses to recommend or independently decide when, where and how much to discount individual products.  SKU marketing agents might even poll their customer persona counterparts described above to forecast likely sales during the next week or month before making a discounting decision. 

Enterprise Procurement: Contract Manager Assistants for every purchase

Contract Managers (CMs) frequently operate at a disadvantage in opening negotiations with a prospective vendor because the internal staff member championing the vendor’s product has failed to provide the CM with the leverage she needs to negotiate the best possible contractual terms.  Champions typically provide an overly optimistic and highly biased view of the vendor’s capabilities, reliability, customer reputation, pricing fairness, contractual flexibility, etc.  Champions are reluctant – either intentionally or inadvertently – to supply any information that might slow down or sidetrack an impending negotiation.

Under these circumstances, the CM needs a dedicated assistant (AI agent) that can quickly compile the information needed to offset the biased perspective of the product champion.  In short, the CM needs negotiating leverage.  Useful information might include such things as product complaints posted on social networks; pending legal actions by a vendor’s existing customers; a list of nearby customers whose procurement groups might be contacted regarding their experiences with the vendor; a detailed review of the contractual terms and conditions offered by the vendor and their similarity/dissimilarity to the terms present in existing company contracts; common pricing practices employed by a vendor’s competitors; etc.  CMs don’t have sufficient time or ready access to the resources required to compile such information.  An AI Contract Manager Assistant does.

Customer Experience: Personalized Onboarding Assistants

B2B companies are acutely aware of the ‘crisis of engagement’ that occurs when end users experience a new product or service for the first time.  New users typically have high expectations regarding the capabilities and utility of new products.  Those expectations always require some adjustment as individual employees learn how to use the product and determine its applicability to their needs and job responsibilities.  Initial engagement experiences can have a major impact – either positive or negative – on the near term satisfaction and long term retention of a new B2B customer.

AI agents are ideally suited to serve as dedicated coaches for individual end users working with a product for the first time.  Onboarding Assistant Agents assigned to individual employees can provide a ‘buddy system’ of personalized super users, providing just the right suggestion or just the right information at just the right time to introduce the new user to a product’s capabilities.  Such agents could provide assistance proactively, without necessarily being prompted by the end user.  They might even collaborate with one another and suggest best usage practices that have been adopted by other members of an individual’s work team. 

Personalized onboarding assistants would eliminate the need for training lectures, demo labs and CBT courses.  Individual employees would learn how to use a new product at their own pace and, most importantly, in the context of performing their specific jobs.  Fear of asking a ‘dumb question’ in the presence of one’s human peers would also be permanently eliminated!

The real limiting factor in AI exploitation might be human imagination

The automation focus of current AI adoption initiatives is somewhat predictable.  Every wave of new technology that has appeared over the past 30 years – from ERP systems to niche SaaS applications to no-code/low-code routines to RPA bots – has been used to automate different aspects of routine work processes.  None of these technologies has forced us to think in a more fundamental way about new work practices that are freed from the historical resource and intellectual constraints that have defined modern business operations. 

In principle, AI should enable us to imagine new work practices, not simply better, faster, cheaper ways of performing tasks we’ve become accustomed to in the past.  The agentic workplace vision that has seized our collective consciousness doesn’t simply challenge the role of human knowledge workers in the future – it’s also a test of our human imagination – our ability to conceive of new work practices that leverage the unique capabilities of agentic AI technology.    

This article was Co-Authored with Shuo Chen, a General Partner at the venture capital firm IOVC and a Lecturer at Stanford University.