"The Best Way To Predict The Future Is To Invent It -- the best method that forecasts future creates future namely.
Understand a technology as natural language most realize the domain that commercializes be born first, industry of intelligent customer service attracted numerous market player to argue photograph position. From manufacturer of traditional call center, to company of SaaS cloud customer service, go to company of customer service robot again, of all kinds company is using AI actively to be endowed with for customer service can. The article basically chats from product perspective, direction of product of intelligent customer service, 2 likely hole and 2 outlets.
Direction of product of one intelligence customer service, 2 likely " hole "
1."System of general intelligence customer service " , probable be " large hole "
Look from the result, it is technology or product side no matter, at present level is the be born that cannot support general intelligence customer service, but, the exploration of this part, itself also is beneficial, the link that does not circle the past even;
2."The solution of intelligent customer service of perpendicular industry " , also have many hole.
If value locates returning is " (for perpendicular industry) cost of economic customer service " , although the likelihood supports a group, but can be mode of a Hard; Because often can be in to be a few big clients to make an item by confine, once was not handled good below a few cage that do not circle the past, also can fall into hole probably in.
(1) AI group itself does not have this perpendicular industry experience, make truly good solution very hard. A lot of moment, AI group needs interior of company of thorough big client to understand business flow, detail, problem to wait; But the in-house stuff that great majority has old experience, oneself comb not clear problem...
(2) retreat one pace to tell, if comb the problem,was clear about, may appear again a new adverse case: If just solve the core problem of 20% , not certain even need uses true AI solution...
And, have truly great " reduce customer service cost " the company of demand, build a group very likely to do oneself. Have cost consideration already, also data security considers. Drip for instance, what a lot of people did not think of is, drop drop has been the company with the biggest cost of domestic artificial customer service, and relevant data very sensitive, so drop drop has huge motivation to establish group of intelligent customer service oneself. Still have A in, also be the time that spent a few years, get on the net shop it is clear that the problem is combed related the customer service that waits for a domain, spent a lot of people to manage carefully and nearly a large number of available data, next just one step by step an intelligent customer service is real of be born.
(3) remove one condition again, for the moment not kink " AI plan " (namely, the problem that assumes the plan of each company can solve a client partly) , at this moment, when returning meeting to face contest bid " make an impact " problem.
(4) hypothesis is very lucky, took the form of a certain big client, how does that follow-up do? The big client company inside perpendicular industry with respect to so much, of company income add fast with gross, can have the ceiling. And approximate outside of the bag make an item, still meaning cost also increase in linear.
(5) if big company is bad to do, does then Xiaogong manage? The medium and small businesses of general China, although feel " reduce cost of artificial customer service " it is meddlesome, but greater power still is in " how to increase income " , namely, how to pull new, how to rise change rate.
Direction of product of 2 intelligence customer service, 2 likely " outlet "
For general orientation, at present this phase becomes AI fall to the ground, "Displace a person directly " it is breakneck (experience from the product no matter, or from commercial feasibility for) , proposal more considers how " AI is auxiliary and artificial " .
Below, from 2B (the company to the enterprise) with 2C (the company to consumer) two direction analysis.
1.2B (TO Business)
Cite a case: Everybody knows to Siri this is planted speech assistant product, adult goes and she is alternant, ask 10 times, although was opposite 9 times, should reply 1 times only bad, so after a month, user slowly again also need not. - - that is to say, from pure technology angle for, although had accomplished 90 minutes, but will look in product angle, the user value of this mode is not worth 60 minutes; If consider commercial value again, the likelihood was not worth 30 minutes (showing level at least is such, future perhaps has an opportunity) . - - that is to say, technical dimension is spent + product dimension is spent, not be to making addition, doing however " normalization " , be the Baseline that cannot regard integral product value as 90 minutes.
2.2C (TO Consumer)
Be based on condition of existing technology of intelligent client product, how to use average user? Have positive and negative two example.
(1) unsuccessful Magic mode (AI+HI)
Before a few years, the Magic pattern that has special fire of Copy of company of a batch of China (AI+HI) , will look at present not too successful, main reason may have:
1.The be in a dilemma that demand locates (head of long end Vs) . To high frequency demand, user itself can be solved with App; And if want to cover too much field length end,mean, need the professional customer service of more fractionize domain; And " customer service of major of more fractionize domain " investment yields comparing is wasteful.
The issue of train of thought that 2. becomes MVP test and verify (note: MVP, the abbreviate of Minimum Viable Product, meaning for " product of the smallest feasibility " ) .
Namely, the logic of this backside is: If choose and employ persons is versed in, can have the effect that 80~85 divides at the beginning, but if use AI, at the beginning only can 60~70 divides the effect; So, if labour of choose and employ persons does MVP, the user is used do not rise, do not have technology of AI of necessary large investment; But if use AI implementation at the beginning, do not have probably as a result of the effect enough good, and cannot judge MVP plan itself whether OK, be given possibly even to procrastinate dead by cycle of AI research and development and cost.
(2) once the Google Duplex of special fire
Google is in Google Assistant of its individual assistant, added Duplex function newly, can help an user call to the businessman such as restaurant, barber shop communicate, book.
The point with the fine essence here depends on, although apparently AI is to serving to tell an user at general, but the customer service personnel that the mutual boy or girl friend that real AI dialog product experiences is a businessman, and means of the conversational flow of itself of this kind of crowd, language even corpora content is opposite more clear can accuse!
Namely, on the contrary, if AI technology is a service apparently at client company, but the mutual boy or girl friend that end item dialog experiences is average user actually, so, to showing skill of level AI product for, flow of this kind of dialog cannot accuse very much.
Brief summary:
Intelligent customer service, it is very good enter travel AI direction. Because relevant company optional limits is not little, relevant position type high school is small have (division of training of AI product manager, artificial intelligence, data is tagged etc) , interview doorsill is done not have so tall (have certain knowledge to NLP, or, if have perpendicular industry experience, can more organic meeting) .
Do you feel what to amused intelligence science and technology still have? Greeting lower part leaves a message comment.
