The essential guide to 'bots' for sales and customer service
This guide is for business leaders and decision makers considering options to automate sales or customer service activities with AI software. You will find an overview of the technologies, their suitability for various situations and the range of results you should expect.
The disruptive change is forecast to be significant: OECD estimates 35% on routine cognitive jobs. Opportunities are for automation of existing work and also undertaking new tasks hard for human workers.
Artificial Intelligence (AI) is a name for a machine taking a series of learnt facts or rules and applying these to a situation to define an outcome. Natural Language Processing (NLP) ascribes an intent to words used. All processing methods use rules or algorithms to define a course of action: the rules may be applied directly to a situation or at increasing remoteness until a probabilistic pattern match is reached. The 'closer' the rule is applied to the interaction the more precisely the response matches the instruction, but the greater the programming overhead.
In a nutshell, if you are realistic about what you are trying to achieve and how you are going to do it, then there is some good stuff out there now. If you want all singing all dancing you will probably be disappointed.
The opportunity to automate repetitive tasks is self-evident. Bots also provide an opportunity to trade in different ways and enter new markets. Scalability and availability allows a conversation with every customer whenever it may be helpful, potentially making self-service redundant. Adding language localisation to 24/7 availability supports global traders.
5. Business readiness
The fastest to achieve results will be those that have an existing demand for assistance and structured processes that can be automated. The choice of technology and delivery partner is, as always, critical.
6. Main choices
For most businesses given today’s technologies, prompt based systems are likely better than NLP; rules orientation is more versatile than probabilistic matching; starting with libraries of rules and dialogue gives a good head start compared to going fully bespoke.
7. Likely winners
The big gains will come for those that can drive greater sales. Scalability and availability enables wide customer engagement. Delivery of defined sales processes devised for specific product areas using prompt based approaches gives an engaging but highly predictable enhancement to current procurement practices.
Do things more efficiently... plus do new things
Gain influence right at the point of purchase
Understand the benefits and pitfalls of recruiting a bot
In a connected world, a hot media topic can garner so much momentum that in a matter of weeks it can climb to the peak of hype and become ‘factual’ in an almost unstoppable fashion. But temper your scepticism that Artificial Intelligence is just another fad. AI is far from new – in fact it has been conceptually relevant certainly as long as computers have been beyond the first prototypes.
The realistically achievable possibilities have become so wild that some of the greatest academic and political minds consider with seriousness the need for safeguards and regulation equivalent to say genetic modification and other work into the creation of life itself. Others focus themselves on its application such as the Leverhulme Centre for Future Intelligence – a collaboration between four of the world’s top 10 universities: Oxford, Cambridge, Imperial and Berkeley.
The Government Office for Science published its latest report entitled Artificial Intelligence: an overview for policy makers on 9 November 2016 considering three questions: What is artificial intelligence and how is it being used?; What benefits is it likely to bring for society and for government?; How do we manage any ethical and legal risks?
This area of technology brings two huge business opportunities – the opportunity to do the same or similar things but much more efficiently and secondly the ability to do new things. The bot - the machine that undertakes an activity in a human like way - has become the medium through which such things are delivered.
Once you have described how to go about a task, the machine can be left to do it either ‘supervised’ to achieve a defined outcome using a process and data set seeded with ‘correct answers’, or ‘unsupervised’ following a base method but with no ‘correct outcome’ e.g. to find patterns.
Want evidence of its significance? The OECD forecasts 35% of employment will change as a result. The area most affected: routine cognitive jobs – white collar workers paid to do rather than decide.
We examine what, why, where, how and when you should be looking at bots as part of your business strategy to grow sales and profits digitally. This is non-technical: the focus is the commercial reasons to act now or to postpone; answering questions such as:
- What types of technology exist and what and who are they for?
- Is the technology fit for purpose for business use?
- Is a bot right for your company?
- Can you realise the benefits?
- How do you choose?
- What are the implementation steps?
- What outcome should you expect?
At the end you should have a good understanding of the benefits and pitfalls of recruiting a bot for sales into your business.
AI uses algorithms reflecting how to do something without specifying every circumstance
NLP derives semantic meaning or intent from words used
Prompt based systems are typically more accurate similar based NLP systems
"Conversational commerce" reflects its delivery through chat based media
"Mobile" describes the device used to access the content.
Probabilistic approaches match patterns to find the most frequent answer
Question:Answer pairs are little more than FAQ searches
Decision trees can be very laborious to train
Method rules provide rapid training and more precise outputs
Self optimising method rules can be highly efficient even with low data volumes
Connected windows or overlay deliver advanced functions
Current technologies: an overview
Definitions vary, and rather than getting caught up in semantics, I use the common descriptions to reflect the broad areas.
Artificial Intelligence (AI)
AI is a term to reflect a series of instructions of how to go about something and the end goal, without specifying what to do in every circumstance. The instructions or algorithms are applied to a specific data set to achieve an outcome. A parallel from childhood education could be considered the difference between responding through ‘understanding’ of a subject area (AI) and learning an answer by rote (rules based responses).
In the context of sales bots, AI represents a means of responding within a conversational message flow (or dialogue) that is either an addition to or an alternative for search and filter: intended to give a more focused or complete response to a question or requirement.
Some purists consider certain bots are not AI, but remember AI is merely the name for a machine taking a series of learnt facts or rules and applying these to a situation to define an outcome. What matters is how suitable the AI engine is for the purpose to which it is to be put. There is no better or worse AI engine, but there are more or less suitable ones for a given purpose.
Natural Language Processing (NLP)
NLP is a mechanism of taking the actual words used by a person and ascribing a defined semantic meaning or intent to the words, which may or may not be a literal interpretation. The first key benefit of NLP is that its use should feel natural – use your own words to describe what you want. The second key benefit is that it can be used in many situations – in fact pretty much any form of verbal media – but most typically in written messaging media such as instant messaging, live chat or email.
NLP is best used for handling a narrow range of requests even though you can ask in many ways. For SMEs this is a really key point – you need to be doing the same thing many times for NLP to really work for you. The alternative: prompt based approaches are undoubtedly more accurate and might be worth considering.
Conversational Commerce (CC)
NLP brings us nicely to CC. The big hype about CC is because consumers are moving. Less time is now spent in apps and more time (now the majority) in messaging platforms. Instead of doing something in an app, or broadcasting using a medium such as Twitter, people are spending much more time in dialogue person to person or in a group. It is therefore important to those that want to get access to some of that time and attention.
Conversational commerce is the ability to reach and sell to a customer whilst in a conversational medium such as messenger or chat.
However, there are some qualifiers as to who should be interested. First up – are you selling direct to consumers or are you selling to businesses. Second – is it really practical to market and sell your product without reference to some form of catalogue or menu?
CC is often confused with mobile. Mobile merely reflects the device that your content is being accessed from. Granted the phone is the ‘first screen’ for personal use for most. But that’s just the point. For personal use. Before you get into mobile, understand the nature of your addressable market, and if it's B2B when is your prime selling time? A user at work in an office for example is better reached through a desktop or laptop. What other applications are they using (including browser accessed, cloud based services)?
Let's move on to the approach of finding an answer (or even deciphering the question). Unless you have very simple needs, be cautious of simple question/answer technologies, which are little more than FAQ responders. Some systems combine several question:response pairs but process them independently. Whilst apparently more sophisticated, these selectors tend to rigidly funnel an answer, ignoring ambiguity, even if they allow individual responses to be changed. Combining multiple question:response pairs before answering means you can deal with more complex queries and provide advice. This is the equivalent of a salesperson gathering all of the customer requirements before proceeding to the next stage of a sale.
Whether the system deals with one or several inputs, they are then processed to find a response. Probabilistic approaches look to match patterns – identify the input pattern and provide the specified output. They work well when there are many hundreds or thousands of repetitions of very similar transactions. Fed with a history of incidents and either the associated ‘correct responses’ or method to create one, this approach provide the most likely answer. Self driving cars use this method to predict what a human driver would be most likely to do in a certain situation.
If there are insufficient repetitions to be viable, or there is zero tolerance to errors you may be better served looking elsewhere. A Decision Tree could be a good option to pursue. However, you are getting closer to providing a specific response for each specific question. The training could be laborious, but the outcome will be highly predictable – great for regulated industries.
The last main alternative is a happy medium: the Method Rules based system. These are just what they say: teach the method rules and let it handle the enquiry in the most optimised manner drawing upon the history of similar interactions.
Technologies can be delivered as part of a web page, as an unconnected window on a page, or as connected window or overlay. Building into a page means changing each page that accommodates the bot. Using unconnected bots restricts ‘show me’ functions (which is also a weakness with most chat applications). Connected but separate deployment can give richer functionality including concierge style experience.
OECD says 35% of UK jobs are likely to be affected by AI
Start implementing now
Be realistic about what and how you will use the technology
Knowledge base of product, process and language is a key foundation
Is the technology fit for purpose?
Software is notoriously difficult to assess when it comes to state of development. It is almost never fault free, constantly evolving and both demonstrations and case studies avoid difficult real world problems. What you see may not be what you are can reliably achieve in your own organisation.
Generally speaking, for first movers there are two key questions:
- Will adoption generate the level of competitive advantage to warrant the investments?
- How much of a head start will be achieved by starting now versus postponing?
For the early majority (those people that like to implement new technology quickly once it is proven) the key question is more blunt: ‘Does this actually work? Where is the proof?’.
Analyst opinion is consistent. Smart machines will create huge disruption in the way businesses do things. It may be that some of the more glamorous claims of capability are not supported by the current actual state of technology. Still the OECD and a separate report by Deloitte are consistent that 35% of UK jobs are likely to be affected. Routine cognitive jobs are likely to be greatest disturbed by bots.
What is the current state? In a nutshell, if you are realistic about what you are trying to achieve and how you are going to do it, then there is some good stuff out there now. If you want all singing all dancing you will probably be disappointed. Beware the grand claims!
Gartner’s well respected technology hype cycle for 2016 comments:
“Smart machine technologies will be the most disruptive class of technologies over the next 10 years due to radical computational power, near-endless amounts of data, and unprecedented advances in deep neural networks that will allow organizations with smart machine technologies to harness data in order to adapt to new situations and solve problems that no one has encountered previously”
Forrester also reported on this question in October 2016 . The conclusions were mixed. My takeaways from their report were:
- Conversational dialogue is the direction of travel for consumer interaction
- Pure natural language interaction has a way to go before maturity
- Despite the shortcomings, the trends and opportunities entice beginning a bot journey
In business language this technology offers some significant opportunities (and potential threats for the non-believers). Businesses should start implementing now. However, stay pragmatic in terms of what can be achieved given the level of investment of time and money you are willing to make. The technology is good but it will still need to be trained to a greater or lesser extent. Development of the knowledge base of process, product and language is a key foundation for this technology.
Bots are excellent to automate repetitive tasks, and will help you reach customers when people cannot
Available 24x7 and cost effective with tail customers
Avoid bots if you: deal mostly bespoke, serve all customers offline, can't engage your staff, or have not started to digitise
For most, hiring a bot is the best sales hire you could make for your business
Is a bot right for you?
How do you know if Smart Machine technology is right for you? If you are looking for a means to automate repetitive tasks or you are trying to reach your customers in a medium that is difficult to access as a person then the chances are a bot can help.
The topic of automation is familiar to almost everyone: software systems are so common place that to a greater or lesser extent they will already be embedded in your organisation. But don’t confuse automation with pushing work to someone else. Web ordering typically is not automation – it is simply a more effective means of displaying a catalogue to a customer. It also takes advantage of the somewhat incongruous belief that self-service is good: it enables a vendor to push the order entry task to the customer under the guise that it is more convenient, saving the vendor time and money in the process.
This is particularly evident in the SME environment, overhead costs have to be tightly controlled. Outside of niche markets, the ‘big boys’ tend to hold a purchasing price advantage, and so account management is reserved to nurture the top 20% of customers. In this context bots can prove extremely useful to automate valuable parts of the sales process and maximise the opportunity from the tail customers. Gains that can be accessed from assisted self-service include:
Deploying your organisations know how to ensure customers make the best purchases,
Increasing conversion rates with improved persuasion and an active ‘closing process’
Growing basket sizes through cross sales.
Many organisations have already deployed human-to-human "live chat" to try to exploit some of these opportunities, but bots additionally offer the advantages of reducing costs, improving the quality of these activities, and offering a scalable solution.
Live chat brings us onto the subject of reaching customers. Some ‘new’ mediums are difficult to access as a person. Plainly live chat was created to connect with a customer at the point of purchase on a website.
This can be extended by embedding your bot into an app making it more suitable for mobile. Conversational commerce attempts to take the concept to a new level and move the dialogue into a leisure chat medium. Of course the owner of this medium has a revenue opportunity to connect vendors to potential customers chatting about their products – a parallel to paid adverts in a search engine.
So is a bot right for you, your business and most importantly your customers? That question is easiest answered by considering when you should avoid them. In these circumstances I would advise you spend your resources elsewhere:
You sell genuinely custom products and that even the sales process is bespoke
Your customers do not want to interact with you on digital media
Your people resist technology and will not represent it positively to your customers
Your organisation has yet to begin a digitisation journey: catalogues, order processing, communication are still paper based and likely to stay that way
For everyone else a bot is an imperative addition to your business.
Every new employee needs examples of what to do... what training is available?
Bots can be great at linking customer problems with your available solutions
Picking the right technology and the right advisor is absolutely key
Realise the benefit: get in action fast
The most critical parts of the process are the ability to train the bot to undertake meaningful added value work, and to achieve increased levels of influence and customer engagement.
A good place to start is considering how you train new colleagues. Almost every organisation will undertake some form of side by side training - an experienced member of staff showing the new colleague what to do. Some organisations will go further with written notes of how to undertake various activities. A proportion of these will have structured and documented processes, check that the organisation follows them, and build in customer feedback loops to refine and improve (the basic principle of ISO9001).
To achieve results from a bot you need to provide it with enough examples of the situations it will come up against and how to respond. The more structured the existing process the faster benefits can be realised. For those that do not have such a structured knowledge management repository, the activity of documenting and reviewing processes is often valuable.
The second key step of realising the benefit is to get customers to engage with the technology. The positive benefits to the customer are typically ease and precision of product selection. Demonstration of benefits to the customer link directly to recognition of the value of sales advice and the more obvious the benefit the greater the willingness to engage.
Combining these factors, bots show a strong business case for organisations that pursue a quality based business approach. It is magnified where they sell products or solutions where the customer’s perceived needs and the optimum solution are not always immediately connected e.g. for complex products or where a degree of expertise is needed to understand feature/benefit.
The choice of technology does materially affect the pace at which benefits can be realised. When it comes to training there are huge differences between a bot using a probabilistic approach and a rules based approach (see Part Two for an explanation). There are also big differences in the required scope between a prompt based bot and a pure natural language option.
Where existing knowledge management can be translated into learned actions by the bot, very fast progress can be made in rule based systems. In business terms, a well-defined sales process makes it much easier to get up and running quickly.
Pre-existing libraries of process and conversation are both a shortcut but also improve quality as they have already been through multiple stages of refinement.
Realising the benefits is substantially accelerated by working with a partner that brings both technical solutions and business know how.
In summary there are four factors that impact your ability to deliver great results fast:
- An existing demand for sales assistance i.e. you already successfully add value to customers by giving advice on purchases
- You have some structure around how you conduct your business: documented processes and training materials are ideal
- You pick the right technology for the environment and data you have
- You work with a suitably experienced and skilled delivery partner
Increasing sales and/or cutting costs must provide target ROI
Expert advice will help make a choice that is a good fit to your business
Avoid probability and NLP unless you have both huge transaction volumes and the most frequent answer is the best answer for customer and you
You may be surprised how an expert can enable the bot to be trained on what seems impossible
Most organisations will succeed if their choices involve a prompt based system, that utilises rules and existing libraries
How do you choose?
Implementing a bot is like other investments. Impact and ROI are significant decision factors. Look to a bot to deliver sales growth and cost reduction.
Sales growth can come in several ways: higher online conversion rates (increasing influence by delivering advice), greater basket sizes, capability to sell new products, increased customer retention (better service). These are proven results from adding dialogue to a website.
Automation yields cost reduction and increased quality.
Ultimately, a successful bot implementation gives your business consistent high quality sales and service, available 24x7 and scalable for peaks of activity. Successful delivery is dependent on the choices you make.
We recommend as a starting point:
- Be clear and realistic on what you are trying to achieve
- Take advice from an expert that understands your situation and can give tailored advice
- Consider the available technologies based on your specific circumstances and objectives
- Decide on how you will refine and develop to future proof your investments
To help chose your bot, we recommend you also think about these issues:
- Complex situations are made harder if communication is not 100% clear. The method used has a big impact on precision
- The capability of the bot is proportional to the training it is given
- Transaction repetitions impact technology choices
Most business people question the capability of bots to cope in their business. They question how the technology will be able to get to grips with the products in the same way a person can. Perhaps that is because people struggle giving solution advice in more complex areas such as when dealing with:
- Large catalogues
- Many similar products
- A weak relationship between product features and purpose/use of the item
- Several products required to work together to form a solution
By selecting the correct technologies, these difficulties can be minimised.
In addition, when handling complex situations, clear communication is essential. Natural language is inherently less precise than prompt based selection because intent is derived rather than stated.
So when it comes to making a technology choice, NLP and probability will produce a very engaging system, but it is going to be very hard to produce successful bot behaviour for highly complex products requiring judgemental advice.
As regards training – consider whether you can provide many examples – thousands of similar conversations. For all but the most repetitive industries such as insurance, repetition rates are simply too low to utilise predictive technologies. The alternatives are logic or rule based. Particularly in this situation it is still much better to start with an ‘educated bot’. Example libraries of conversations and business based flows give you a big head start. But watch out for toys. It would be remarkable indeed to be up and running in 30 minutes!
Remaining on training: this will be an ongoing process – to extend capability plus to stay up to date and reflect changing circumstances. Training is an absolutely key activity. Check for tools that enable this to be undertaken by non IT staff.
For the ‘average’ business with current technologies we recommend considering:
- Prompt based systems over NLP
- Rules over probabilities
- Existing libraries/out of the box training over fully bespoke
These choices are likely to enable you to be up and running in days rather than months, whilst not overly compromising the actual capabilities of the system in the medium term.
A bot giving good advice builds trust and desire to reciprocate
Customers that buy the best product for their need get better value and are more loyal
Influence at the point of purchase helps close more and higher value sales
Ensure you have clear objectives and a considered plan
The process of bot implementation can help develop your other sales too
Engaging your bot with customers is a key stage
It's not just sales process and advice that your bot can automate
Know where your bot is trying to take you and be sure to celebrate when you arrive!
Let's get this party started....
Influencing customers at the point of purchase helps you to close more sales, and build greater long term satisfaction. Giving good advice will build trust and create value. So adding a bot to your website, and a bot that sells, is a must do task for successful companies in 2017.
So far we have explained what you need to know to make a great choice, now it's time to get the party started and get on with implementation. You need a plan - preferably agreed across multiple functions including sales, marketing, operations, ecommerce and IT.
Here is a quick overview to get you started.
Firstly, ensure you are clear on the scope of the project by asking:
- What categories or products or activities will your bot cover?
- Where should it appear – homepage, category or product pages, every page?
- What is it going to do for you? From receptionist to general manager, account manager to product expert …
- How will it respond to ‘sorry, but I don't understand’ situations?
By agreeing on scope up front, you can be sure that success can be measured in stages, quickly identifying which customer interactions benefit from the bot interactions. This enables further stages to build on these successes, and avoid parts that have little or no interaction.
Secondly, look closely at training and your environment by understanding answers to:
- Are there any regulatory factors that need to be taken into account?
- Is it your intention to replicate an existing process?
- How is training to be done once the commercial flow is defined: specialist IT development task or business focussed "code free" approach?
- Does your product data support differentiation between products?
- How will you manage ongoing activities?
Examining the case for a bot can quicken the understanding, documentation and optimisation opportunities with existing processes. Ensure to ask questions to maximise product differentiation in the areas that matter. Some bot providers will provide a library of content to reduce the need to create everything bespoke. Building a bot is not a one-off job, it requires tuning and adapting to provide an enjoyable experience for customers as well as learn new product categories and facts – it is important that the team defines how change will be managed and approved; as well as any auditing and tracking that needs to be undertaken for compliance.
Thirdly, engagement: how people will "get to know" your bot - and that it is available.
As there is little purpose in having a brilliant bot if it is unused. Ask yourself:
- How will you go about building customer engagement?
- Where will it be visible, and how will it initiate conversations?
- Do you reveal your bot as software up front or pretend it is a human?
- Is it appropriate to "wake up" the bot from other events – such as certain search requests?
- What sort of timing do you want between responses to make it natural as possible?
- Which is more important – what is said or how you say it?
Yet more questions, but the critical fact is that you need to build engagement with the bot to understand how your customers will respond. Even if the interaction seems mildly intrusive, the important thing is getting it in front of people and giving them the option. By tracking both open and reject events, you can then build a picture of which pages are most likely to garner interaction.
Fourth: integration into the business - a bot that works with other systems
Clearly there are going to be a few "technical" aspects on a checklist. I promise you I've cut these down to the bare minimum so we can talk about measuring success shortly! So, it is critical to check with your bot vendor a number of points including:
- Putting the code live: is this a simple drop in script or is there heavy lifting to be done on your ecommerce site?
- Pass to an agent: do you have a robust and timely method of passing a conversation to an agent in a way that will yield increased customer satisfaction?
- Ecommerce actions: Can the bot prepare and place recommended products to basket automatically following a conversation?
Five Alive: Our bot is a great success! We measured it!
- Finally you need to be up front about how you're going to measure and manage a successful outcome from your bot implementation. So you need to check:
- What is the benchmark to measure success: conversations, resolutions, orders?
- What are the targets and how does this link with ROI?
- What metrics can be captured by the platform and what do we actually need?
As all great advisors say "forewarned is forearmed" – starting your bot project with this checklist of questions and advice will give you the best possible chance of success. And of course an amazing party to celebrate success.