Practical steps to transform fragmented customer records into a powerful business resource Do you...
The problem
A large technology company with a wide range of products was entering into new technology segments in response to market shifts.
Customer-facing sales and support staff were less familiar with the new product technologies, which created a bottleneck in pipeline throughput as deals required support from scarce resources in product owner and engineering roles.
Customer experience and sales were negatively impacted by staff inability to answer queries in a timely manner and contention for specialist resources.
The solution
This problem is common when we look through a customer support lens rather than a sales lens.
We want to preserve an excellent customer experience, which in this case would amount to promptly answering detailed questions that arise during the Qualification, Solution Development and Post-Close phases of a typical sales process.
However, we have a constraint on the availability of the most skilled resources to answer those detailed questions. We might hire to grow our resources in this area, but we probably can't completely solve the problem that way as it takes time to hire and ramp. The customers - and the opportunity - are here now.
We thought about this in two ways:
- Pre-emption: How do we reduce the need for the specialised support resources but still provide an excellent customer experience?
- Curation: When a specialist is needed, how do we focus them on the right deals?
Pre-emption
The most strategic accounts received dedicated specialist resources. The overall account list was determined through machine-learning-assisted analysis, however the assignment of dedicated specialists was a human call.
Not everything is a technology solution. In this case, strategic clarity enabled focus. This can be uncomfortable, as it involves an apparent choice about where to withdraw.
In reality this is not the choice it appears to be. If you can't adequately support the whole customer base with current means then you are not really withdrawing. You already have withdrawn... you are just pretending you haven't.
We enabled sales and support staff with new content to address the common queries.
Content Overload
However, there was too much content. This presented non-specialist staff with insurmountable problems:
- How do I know which document to read?
- I don't have time to read the documents
- How do I know if the answer is in there? This was a major problem for customer-facing staff. It takes much longer to prove there are no needles in a haystack than to find a needle in a haystack!
We implemented an enterprise RAG solution to make existing product documentation queryable. At the time this was implemented by a dedicated AI team, but now that capability is integrated into major platforms so your existing IT team should be able to handle it. I would strongly encourage the use of off-the-shelf platforms for AI use cases that are now more routine.
Salesforce describe RAG like this:
Retrieval-augmented generation (RAG) is a natural language processing technique that merges the best of retrieval-based and generative models. Information from a database or knowledge base is used to enhance the context and accuracy of generated text.
RAG allows companies to connect their data with LLMs, enabling artificial intelligence opportunities for businesses that are more trustworthy, pertinent, and timely. For example, once the connection is made to internal data with RAG, autonomous AI agents can deliver customer service responses that take into account past questions or generate marketing briefs based on current brand guidelines.
This was transformational.
- Staff had one interface for all relevant product documentation
- Answers were immediate
- Critically, if the answer was not contained in the documentation then that was apparent immediately, instead of after days of searching around
- Unanswered queries were captured to drive support document improvements
Curation
In the Sales CRM we implemented a request-for-support workflow with the following features:
- Request for support was routed to an empowered review agent based on the product
- Logic on the request workflow checked that key information to enable specialist allocation was captured. This included the customer and deal details in order to prioritise specialists for the best opportunities.
- The support team was added to the opportunity team enabling collaboration and visibility to relevant deal information. This curation of opportunity team membership is critical for federated organizations and customer-sensitive work.
- The agent reviewed and made a resource allocation based on business priority and agent availability. This was human-decision making which could be augmented or replaced by an agentic AI solution. In this case we wanted the human judgement for this particular decision. That was because, given the nature of the new products, there was little objective data on which to train a Machine-Learning-based agent. Given more history of attributes, choices and outcomes then an AI-based classification solution could be considered.
The outcome
These actions measurably improved the throughput time, sales, support and customer experience:
| Action | Outcome |
|
Strategic focus on the most important deals Recognition that not all can be equally important |
Focus on most qualified deals removed distractions |
| Dedicated resources for strategic accounts | No need for strategic customers to wait at all, specialists on hand. Time to engagement reduced to zero |
| Improved content, queryable by chat-style interface | Time to Information reduced from 3 days to minutes |
| Support workflow | Resource allocation decision reduced to within 1 day from 3 days as no time wasted exchanging needed information on the deal |
Key Takeaways
- Start with strategy, not technology. Ensure the business focus is clear
- Focus on removing bottlenecks and then streamlining what remains
- Leverage off-the-shelf capabilities, including simple rules-based workflows as well as AI-based solutions
