The Conversational Gap Model – A Personal Journey to Conversational Commerce

Guest Post by Paul Sweeney

Many years ago a Chilean gentleman named Fernando Flores developed an interaction model that modelled workflows as conversations. At it’s core was the concept that there were four distinct phases to a conversation, but that in general, organisations did not model these phases accurately or with enough granularity. It was the breakdown in the integrity of the conversation between people that led to the breakdown in the business process because we were not being specific enough where we needed to be, and were probably being overly prescriptive where it didn’t matter to the ultimate delivery.

From the diagram above, some of you may recognise this way of thinking from your exposure to The Lean Start Up. It’s basically a methodology for running a discovery process.

I was very ready to be persuaded as to the usefulness of such methodologies after a three year masters in new product development processes and buyer supplier relationships in the automotive industry, a theory based on lean manufacturing and lean enterprise. My research involved disentangling the effects of good internal new product processes from the attributes of buyer supplier relationships. Foremost here was the importance of getting the design specification right up front, but not defined “too tight”, such that you lock out the expertise of the supplier. Many times, you defined the “outcome” such as “this switch must work in environmental temperatures ranging from -40 to +40. How you achieved this as the supplier, was up to you.

See also: Conversational Experience

The relationship, communication and trust factors between the car company and it’s suppliers were then really important in guiding and steering through course of inevitable adjustments and adaptations. Getting to market in a much shorter development cycle time, and then iterating faster than the competition made Toyota a learning machine. The same factors are at play when people collaborate. We need to understand the nature of our relationship, we need a “common background of obviousness” in our communications, we need a history of interaction that has built a base of trust. Only then do we trust to iteration.

CEBP — Communications Enabled Business Processes

In 2003 Graham Brierton and I were co-founding team at a first generation SaaS company. We recognised that automated communications could be used to address the millions of “small interactions” between companies and their customers that could be better managed through more accurate preparation, negotiation, performance, and acceptance routines. With the rest of our team we identified a range of “lags and drags” in business processes, where if they could be handled more efficiently at scale, it would free up organisational resources. We saved companies millions of dollars. We won a global innovation award, national and European customer service awards, but yet, we felt that there were bigger “conversational problems” out there that we had not yet tapped. The problem was that we couldn’t “get to” the granular data that was locked in the enterprise, voice recognition wasn’t good enough at that stage, and there was still a chasm between the communications channels and the emerging, enterprise digital infrastructure. We could see all this potential “white space” in the enterprise, but there was no way of “wiring it all up”.

Martin Geddes — The Ten Conversation Gaps

Around 7 years ago I was at a Future of Voice workshop run by Martin Geddes and Dean Bubley and the concept of “Conversation Gaps” was advanced. Again, this was an example of modelling out a typical process and demonstrating how telecommunication services could be used to remove friction, and help business processes complete more efficiently. I’m going to quote it in full.

  1. The cost gap: The difference in operational cost between human labour and an automated system is several orders of magnitude, and errors introduced by humans cascade into further cost.
  2. The confidentiality gap: Human beings are asked to handle sensitive data, and such data can leak.
  3. The customer experience gap: The customer’s time is wasted on tasks that do not create value. The ‘cost’ of the service to the user is the combination of its price, and the effort to use it; thus poor customer experience is a form of cost shifted on to the customer that reduces demand and willingness to pay.
  4. The capability gap: Our tools of conversation assume a one-size-fits-all and do not provide features reflect the diverse roles people adopt in their daily lives, and the different demands that arise as a result. For instance, a next-generation Caller ID would command different levels of caller information disclosure for each of ‘child’, ‘friend’, ‘customer’, and ‘stranger’.
  5. The co-presence gap. Ideally conversational media would provide parties with an experience as good as ‘being there’ together. If the conversation is locked into a narrow range of media types with weak interactivity, then it falls short of ‘being there’.
  6. The coverage gap: The ‘coverage’ is the range of situations in which the media make the conversation possible at all, despite ‘not being there’ together in the enterprise’s retail outlet. If a customer’s available modality of conversation is Skype, and the enterprise cannot originate or terminate Skype calls, then they cannot converse, and the enterprise loses business.
  7. The capacity gap: Each tool has to be able to scale to meet the needs of enterprise use. This is not just a question of networks scaling to accommodate load, but also of the user experience being able to scale to prioritise, filter and route an ever-rising number of requests for interaction.
  8. The conformance gap: The tools of conversation fail to meet the legal and social norms of each jurisdiction, and thus limit their use in commerce. The current controversy over encrypted BlackBerry messenger use that cannot be intercepted by the governments of the UAE, Saudi Arabia and India is an example of this.
  9. The culture gap: Each society has different norms and historical associations with communications, such as Americans having a higher propensity to use voice compared to text-centric Europeans. What is acceptable and ordinary use of personal data in the USA is regarded as unethical in Germany, even if it conforms to legal constraints. Tools of conversation need to reflect these differences.
  10. The ‘cool’ gap: The other gaps address functional shortcomings of our tools of conversation. However, an increasing proportion of the value we receive from goods and services is provided by non-functional aesthetic and social value. People want to use communications services that make them feel and look good. If teens are tweeting and texting, then asking them to talk might as well be asking them to telex — it just isn’t going to happen.

Digital Conversations

Around 5 years ago, we noticed more widespread adoption of “customer journey”. The digital glue for this journey was going to be “social”, but at the customer level and at the organisation level (employees). Social didn’t turn out to be great for engaging and collaborating across all environments and businesses. Gaps were emerging between online digital and the traditional sets of communications tools at the contact centre. While there is still a great deal of promise in the community model of customer support there is little actual adoption of it.

My Personal Mobile Experiences

For me, three customer experiences seemed to be fascinating me, and seemed important.

(1) Instant, Visual, Services: Hailo was an app that I used a lot. I opened it, it found my location, showed me little icons of cabs, showed me arrival times. I hit one button, and I had booked a cab. The cab drivers were unfailingly polite, the journey route and choices were always explained to me in advance, and the bill was always so easy to deal with. I just stepped out of my car and caught my train. Why weren’t digital interactions with all companies this simple I thought?

(2) Private Social Networks: WhatsApp: I was speaking with a long time friend and former colleague who was running a small service company, entirely on WhatsApp. What I said? yeah, we open up different groups for each subject. Basically he was nearly able to run his business, and his customer interactions, from a mobile messenger app. combined with what I was already witnessing with the adoption of Slack, I could see a new assisted interaction paradigm emerging.

(3) Customer Empowerment: I was in a hospital and found that I had absolutely no record of my conversation, and no way to capture the record of my interaction with that hospital. There were many errors made, and the hospital was the only party to the interaction that was “allowed to keep the official record”. Why?

Conversational Commerce

The bridge between your communications and digital infrastructures turns out to be the Messaging app on your phone. If you could @ message a company and have them respond over Messenger, you save a lot of time, effort and frustration. I don’t have to look up my flight information because KLM not only shares my ticket with me in my Messenger, but also my updated gate number, and the time it will take me to get to my gate. The conversation is the service. When I want an Uber it can be summoned from within my Messenger. When I want a receipt it can be shared into our messenger conversation. When I want to remember what was said or agreed, I can scroll back in our conversation. I can absolutely see that in many professional services interactions I would want to keep a record of the preparation, the negotiation, the performance, and the acceptance stages of an interaction.

What we are sure about at Webio is that the future of interaction will use all the capabilities of mobile, from icons and emoji, to micro-sites and bio-metrics. It will us the camera and it will use augmented reality. It will use intelligent messaging, messenger apps, disposable apps, instant apps, sms, webrtc voice and video. The availability of this channels “everywhere” gives companies the ability to bridge many of the conversation gaps originally proposed by Martin Geddes.

Powerful Artificial Intelligence (A.I.) engines such as Tensorflow are also coming on stream at the same time as all these channels, bringing the possibility of predictive interaction to conversational interfaces. For simple, easy to formulate questions, for which there are simple, structured, easy to present response options conversational interfaces should be easy to implement. Where there are more complicated questions, that requires data from numerous enterprise applications, the ability to generate these response options is more difficult.

But, What’s So Interesting About This, Really?

Chatbots are the trojan horse for artificial intelligence products. They are really augmented conversations. They will start simple and small in places like customer service. What devices will do for us “automatically” is only in it’s infancy. Using image recognition on x-rays is one of the areas where A.I. assists doctors and behind that is the fact that A.I. is capable of keeping the alternative diagnoses “in mind”, their potential interdependencies, and their risk weightings. While human experts may be right 95–99% of the time, A.I. is helping chase down that last 1–5% and weighting it for potential impact. If A.I. is assisting doctors in diagnosis, and assisting drivers to remain safe on the road, it is hard to think of an area that will remain untouched by A.I.

In 10 years, machines that help us keep these options in mind, and help us to better exercise our judgement will be the norm. Today we think that Google Assistant is pretty good at recognizing our voice search requests. We mostly think Siri is funny but not as good as Google. We really haven’t seen mixed reality solutions such as Hololens used in our day to day lives yet, but can you imagine going back to a world without Google Maps? Could the whole “on demand Economy” work without Google Maps? At the moment not really.

The conversational interface will be like maps of conversations. At first it helps us get from A. To B. Future generations of services will be built upon this this basic metaphor. If maps were about locations, their killer app was “step by step turn instructions”. It was about the navigation. So if our conversational interfaces are mostly about our “intents” what services will emerge based on these intents? What is our turn by turn navigation equivalent? What is the killer app for augmented conversations? For me, it’s personalization. Conversations are where we explore, create, define, and share our preferences. While these may start out being functional and factual preferences such as remembering our shoe size, and our color palette preferences, they may evolve into preferences that surprise us, that delight us. Spotify can play your favorite album all day long, but all anyone talks about is how amazing their Discovery Weekly playlist is. How many of us get up on a Monday morning and hit that first thing? It’s now “a thing” that many of us do on a Monday. What if we had our “banking conversation” on the first Monday of the month? What if we had a target of having at least “four hours of direct conversation with my child” every week? What would happen if there was some way to know that “25% of people you work with don’t seem to want to communicate with you”?

In an era of quantified communications, we will have augmented conversations, assisted by artificial intelligence that holds a repository of our preferences.


Making Returns on the Conversational Economy

Article by Adam Rawot, CEO Woveon

I remember reading an article almost ten years ago talking about how teens were sending over 40 texts a day on average. The tone of the article was incredulous, but the statistic pales in comparison to how we exist online now. Speaking personally, it’s not implausible I send off 40 messages before 10 AM in my morning inbox check in. Sarah Guo, a partner of Greylock, expressed it succinctly when she took to Medium: “More than a decade ago, academics such as Thurlow described a “communication imperative”—human beings are driven to maximize their communication volume and satisfaction. More recently, researchers have described it as compulsion.”

While constant connectedness is old news, technology has finally achieved a point it can leverage this behavior. As with all big shifts, there will be survivors and those who don’t adapt fast enough. Companies will need to change to a conversational mode of thought to maintain the experiences users expect and deliver the individuality anticipated.

People Always Talk

multichannel communication

Nearly 25 years ago, Harvard Business Review wrote “today if you’re not on the phone or talking with colleagues and customers, chances are you’ll hear, “Start talking and get to work!” In the new economy, conversations are the most important form of work.” Conversations are how we track knowledge flows. Conversation flows are how people create value, share information, and illustrate how companies operate.

A cited example is McKinsey. McKinsey prides itself heavily on the intelligence of its members, and by an extension the true value of McKinsey over other firms is its extensive knowledge base. That knowledge is curated and developed through internal conversation and shared through internal conversation. In short, McKinsey is conversation.  

We are entering a new age for product development – one dictated by the conversational economy. Broadly, the conversational economy is the catchall phrase for companies, products, and ideas built on, alongside, or relying heavily on a conversational interface. More simply, they are services that leverage conversation.

This definition is board, and intentionally so. While some apps like iMessage, Snapchat and email obviously fit into this definition, conversation works as a backbone in services like Facebook, customer service complaints, and online advertising as well. Finding a common backbone helps derive a working model for these services.

Between the myriad of mobile apps used every day, access to the internet, and the seemingly innate human need to feel connected, conversation based platforms are dominating our lives. We have effectively destroyed the asynchronous quality of day to day life. We persist online, and, consequently, our conversations with one another never really begin or end. This data stream is a jackpot for product creation.

Smarter Everyday

Artificial intelligence, in the eyes of the public, has snuck past an important threshold. Presentations by titans like Facebook and Google have assured that we are moving away from the robotic idea of natural language processing in a rigid sense to natural language understanding. In other words, instead of responding to a keyword or a phrase, computers are beginning to be able to understand sentence, paragraphs, and intent.

There are a variety of causes for this – improvement of machine learning and deep learning, Moore’s Law, and rate of mobile and app data collection, to name a few. Algorithms and software are taking on their own intelligence. Just the idea that failed outcomes can make systems better is an astounding twist compared to five years prior.

Additionally, we’re in the middle of the boom of ambient computing, the idea that our environments and surroundings are responsive. We don’t have to open our phone or flip open a laptop to be connected. On the way to work I may pass a few smart cameras, a plethora of listening iPhones and Galaxy phones, an Alexa, Chromecasts, and more. Despite this, I would characterize myself as one of the less connected people in my demographic. At every step of my day my voice can be heard, position tracked, and activity monitored. Being connected no longer has much to do with if our phone is on our person or if we’re behind a keyboard.

Although passive collection has subtly pushed past our natural aversion to share information with technologies we don’t understand and people we don’t know, this one-sided trade has come with the expectation of usability. When software doesn’t work or apps crash, we no longer blame ourselves, we blame companies. We are inundated with choices, but that means that we have little tolerance of poor experiences. Users are more empowered than ever in that they don’t have to subject themselves to experiences they don’t want or content they’d rather avoid. We so demand these freedoms that events like net neutrality rapidly cause public outcry and social faux pas by companies like EA tank sales.

Computing, connectedness, and data almost completely undermine how product managers need to think about designing products. The need to leverage conversation to deliver value has emerged as one of the most critical company problems. IDEO acquiring a data analytics company, giants like Apple acqui-hiring boutique companies with human-centric software, and Salesforce pushing Einstein all serve as mine canaries that even the most established companies are racing and struggling to adapt.

Buying In and Cashing Out

As George Box famously cited – all models are wrong, but some are useful. Where is the utility of viewing products as ongoing conversations?

A helpful place to start is in how companies have historically fended off competition. These ‘moats’ include things like brand loyalty, unique data sources, and intellectual property. However, as technologies like AI are more readily available via open source projects, cloud hosting and computing are only a few clicks away, and systems of engagement continually emerge, the traditional ideas of tech defensibility are evaporating. In a Greylock article on Medium, they wrote “In all of these markets, the battle is moving from the old moats, the sources of the data, to the new moats, what you do with the data.”

In another words, companies are now finding defensibility through the experiences they create. To create these experiences for customers in the conversational era, companies will have to harness existing behavior, respond personally, and work faster.

Harnessing existing behavior is an exercise in invisibility. The real frontier for conversational companies to generate solutions for problems before the consumer is even aware. For example, Facebook realized that people asked for recommendations on their newsfeeds. Instead of creating a new service, they had posts automatically update with reviews and locations. They created a new card that changed automatically depending on what a user wrote. As expressed by a product designer at Facebook: “We didn’t try to invent a completely new behavior; rather, we found an existing one and made it way better.”

To cite an example within my own career, food industry companies often lose hours if not days within food recall investigations. Tracking a faulty shipment through several distributors can be tricky. We worked to create a product that reads the complaint before the owner may even be aware it exists and start and investigation. By the time an owner is even aware there is a problem, a report is ready. By approaching complaints, invoices, and shipments and messages between companies, value can be created seamlessly in a second layer.

As I’ve written about before, personalization is an increasingly critical element of producing customer lifetime value. Harvard Business Review started to notice this trend in their research on customer service: “Even as artificial intelligence becomes embedded in everyday interactions; human conversation remains the primary way people make complex purchases or emotional decisions.” The fatal error in a lot of software products is focusing on company efficiency over consumer experience. While these changes may boost bottom line in the short term, they encourage competitor entry and consumer drop off.

Conversational apps have an obvious outlet for personalization, and the power behind them allow easy switching between automation and human elements. More simply: “these intelligent agents will facilitate one-on-one conversations between consumers and sales or customer service representatives rather than simply replacing human interaction.” Imagine a case where someone sits on a delayed flight and sends out an angry tweet. A conversational built system could find the message, tag it, and route it to an agent. While the agent delivers a personal response with an update, the system has already sent an alleviating reward of extra miles to the customer. The captain may be alerted of sentiment on the plane and deliver an announcement. While an autoresponder may have been cheaper, the customer will now remember the exceptional level of immediate service and is more likely to return. As information and computing become free, the real commodity becomes the personality of the person on the other end of the line.

In the shorter term, there’s a simpler way to think about AI adoption – people don’t trust what they don’t understand. In the classic product management advice, it’s best to start with a problem and move to solution. Leveraging conversation is a means to building a better product, but that doesn’t change what the bottom lines should be. In other words, “Bots do not need to be human to be human centered.”

Outside of the shift in new product priorities, another major implication is how we use the technologies we use currently. In a blog post, Dan Rover (sp?) declared that bot won’t replace apps, but inboxes were the new home screen. Our email, text messages, and more were queues demanding our intention and driving our usage.

Companies leveraging platforms like WeChat have been able to effectively create micro services and apps for things like ordering that have integrated seamlessly with how we act now. Bot companies that are able to daisy-chain onto conversations to do scheduling and commuter planning have shone in venture capital funding. It’s not inconceivable the next unicorn will have nothing to do with creating a new platform but layering effortlessly onto the ways we talk with those platforms now.

Speak Now

We talk online all the time, but computing has finally let us create value from that. Companies need to invest in ways to leverage these conversations to deliver seamless and personal content. This means focusing on personnel and focusing on alleviating frictions than automation. Companies that don’t value the communication imperative and connectedness of customers will soon find themselves lagging in experience, and, later, sales.


A prime example of this is Amazon Web Services’ fast climb to dominance. Legacy systems like Oracle required costly deployments and developers, and setting up cloud instances on AWS is only a few clicks away. IaaS records have shown Amazon’s sheer dominance. Oracle, trying to defend by housing data and curating an elite brand, couldn’t compete with Amazon’s engagement accessibility.

Perhaps the most obvious implication of smart conversational apps is efficiency. However, despite all the news and hype around an artificial intelligence singularity, businesses – and their customers – still revolve around the interactions person to person. This means that products needs to be resolved around facilitating conversation, collecting information, and iterating form that information. The AI boom has made it easier than ever to facilitate personal conversations no matter where customers are online.