- All right well, good morning everyone.It's good to see somefamiliar faces in the audienceand lots of new faces as well.Welcome to our session today,resolving customer issues fasterwith the Amazon Connect agent workspace.If you are in this room right now,you probably operate a contact center.Maybe you're already using Amazon Connector you're considering Amazon Connectfor your contact center.You are also likelyfacing some challenges.Perhaps you maintainsoftware infrastructurethat your agents currentlyrely on to help customers,and you're looking for alternativesto drive agent efficiency.Well, we're excited to be up here todayand tell you about a new launchthat was announced yesterdayin Adam Selipsky's keynote addressthat we think will help solvesome of these challengeswith cloud-based technologyand intuitive agent interfaces.I'm Sandeep Rao, go-to market specialistwith AWS productivity applications.And this is Jack Hutton,senior product manager for Amazon Connect.Let's get started.All right, so many of youalready know what Amazon Connect is.For those of you who do not,Amazon Connect is a truecloud-based contact centerthat allows you to scaleup to any size workloadand you only pay for what you use.It's very simple to get started.Anyone from a non-technicalbusiness leaderto experienced contact center veteranscan begin innovating onbehalf of their customerswith our self-servicegraphical user interface.You can design conversational interactionsthat are natural and personalusing our speech recognitionand natural languageunderstanding technology,the same technology that powers Alexa.And the combination of these benefitsallows organizations todesign customer experiencesthat meet expectations,drive business goals,and reduce costs.- Connect is one of thefastest growing servicesin AWS history.Tens of thousands of customers use Connectcreating over 10 millioncontacts in total per day.Large customers such asNational Australian Bank,Adidas and Delta lovethat they can experimentand move at a rapid pace like a startup.Meanwhile, smaller customerssuch as Answerconnect,Slice Pizza and Chartspanlove that they can getan enterprise gradesolution at startup prices.Overall, we value our customer's input,and it's very important thatwe listen to customer feedbackin order for us to innovate and allow youto rapidly expand yourcustomer service operations.And so I just wanna give aquick thank you to all of youand all our customers for trusting usas you overall transformyour customer experience.- All right, so you'veseen some organizationswho use Amazon Connect today,but how do agents use Amazon Connect?Well today agents can useour free out of the boxagent workspace to getwork done more efficiently.This agent workspace isavailable virtually anywherevia web and single sign-on.All of our agent empowerment capabilitiesare unified in a single pane of glass,but again, you only pay for what you use.The interface is intuitive,so agents can quickly andeasily find the informationthey need to resolve issues.If we dive deep into the agent workspace,we can look at a fewof these capabilities.On the left hand side of this screen,you'll see the contactcontrol panel where agentscan receive voice contacts,chat contacts and tasks.An agent can also see if a user has been,or customer has been, authenticatedusing our voice biometric solution.In the center of the workspaceis our customer profile.Here, customer information is aggregatedfrom disparate sources anddisplayed to the agent.Information like prior contact history,basic customer information,like name and phone number,as well as other additionaland custom informationthe agent or the organization enters.We also have another tab called Cases.With Cases, organizationsand customers can track,collaborate on and resolve issuesthat span multiple contacts.In the Cases tab all the cases associatedwith the customer profile are listedand an agent can click into the caseor create a new case based on the issue.In the top right of this workspace,we have automatic recommendation service.Amazon Connect can listento the conversationand automatically suggestknowledge articlesto help the agentresolve the issue faster,or the agent can type akeyword into the manual searchto pull up a relevant knowledge article.I just want to emphasize againthat this agent workspaceis free out of the box andit organizes and unifiesall of our agent empowerment capabilities,which you can use ona pay as you go basis.- So now that we have an agent workspace,we did ask ourselves whatcan we do more to helpand to identify three key areaswe wanted to help customersmake improvements.The first has to do with onboardingand ramping up new agents andmaking them proficient faster.The next has to do with making surethat all agents, experienced or novice,are adhering to best practicesand following updatedstandard operating procedures.And lastly, we wanna helpbusiness operations teamsfigure out new ways toimprove efficienciesand make sure their overallagents are operating welland adjusting to new use cases.And so with that, as of yesterday,we've launched step by stepguides as a preview featurein our Amazon Connect workspace.And so with this feature you're now ableto configure the primarytab of the agent workspace,so at the start of the call,you can surface relevant info and actionsto your agents to resolve issues faster.At the highest of levelswhat we have done iswe've given you the abilityto surface contextuallyrelevant data and actions,making sure the agent hasall the information they needto get going quickly andthe tools to dive in fast.In particular when it comesa little bit down deeperto a step-by-step guide,here is where you're ableto really structure your SOPand codify into the system so an agentis given clear instructionsof what they should be doingat any point in time when trying to handlea particular use case.So in terms of use cases,it's gonna vary greatly bycustomer and line of business,but at the highest level,we do see four common stepsthat all contacts have,and we really try to optimize the systemto make these steps work well.The first has to do withidentifying the customerand making sure they are who they are.And so with that, wehave customer profilesand voice ID plugins thatmake it a little easierto get this step done quicklyso the agent can dive fastinto the actual issue.But first they gotta actuallydetermine what that issue is.And so with that in ourlanding page experiencewe give you the tools youneed to help guide the agentto figure out what'sgoing on in this contactso that they can find the right actionto carry forth and solve it.Now when it comes to resolving the issues,which is probably the heartof our step by step guides feature,there will be various differentapproaches you can take.And later on in the deck we'll dive deeperinto how this could work.And lastly, most contacts requiresome type of after contact work,whether it's call notesor distribution codes.And so with the guidedexperience we provided,you'll be able to configurethis in our agent workspace.- All right, so let's see,step by step guides in action.The video I'm about to show you,there will be an agentworking at an airlinewho uses guides to help a customerwith a lost baggage reimbursement claim.- [Voiceover] In this video, we will showhow the agent will submita claim for lost luggageon behalf of a customer.When a contact comes in, wewill display the match cards,which is a quick snapshotof the step-by-step guidesthat were configured by managers.When agents click on a match card,we will display the solve cardswith actions that willinitiate a workflow.The customer informationis already pre-populatedfrom the Amazon Connect customer profile.The most recent flights areavailable, and once selected,the rest of the information is populated.The agent asks the customerwhat items were lostand the amounts that the customerwishes to get reimbursed.The receipts can then besent via email if needed.In this step, the customer agreeswith the reimbursement amount.Finally, the agent can reviewall steps with customersto make sure all of theinformation is accurate.And the form is successfully submitted.- [Sandeep] All right, welllet's call out a few things.The screenshot you see hereis step one of the guidefrom the video we just watched.Now, guides open up in thesame tab as the landing page,so an agent doesn'thave to leave a contact.Further, there is a progress baron the left hand side of the guideand basic contact detailson the top of the guide,so the agent never losescontext on the contact.Information from first party sources,like customer profiles, canbe used to pre-populate forms.Information from third party sources,in this case, flight management systems,can also be brought into the guide.There are also navigationbuttons at the bottom,so the agent can progressthrough the guideor cancel the guide altogether.Now you've seen an exampleof how guides work,and you're probably asking,why have we designed guidesin this specific way?- And so we wanna take a small segueto explain how we've used guidesin our own Amazon customerservice departmentsin order to improve agentefficiency and training.And so our old customCCP looked like this,and as you can maybe tell,there's a lot of features for agents,which is great because we gave themall the power they needed toresolve any type of contact.But as time went on andwe added more features,and use cases got more complex,this started to become a burden on agentsbecause it became a veryunwieldy system to use.So for example, just to look upthe status of a Primesubscription membership,agents would have togo through three menusand four UI pages just toget the status details.And then to actuallymodify the subscription,they'd have to do that three more timesgoing through 15 differentUI pages and many menu tabs.And overall this is becomingtoo burdensome of an experiencethat is placing highcognitive load on our agentsthat we've noticed was reducingtheir ability to learn.And this is true fornew agents just tryingto learn the system and getacquainted with their job.But even experienced agents we realizedwere diverging in how they handledthe same type of customer requestsbecause the system was so fluid and open,and that was leading toinconsistent customer experiences.So we dug deep and we tried to evaluatewhat is really going on with our UX.And when we evaluated it,we realized our UX processwas asking the agent tolisten to the customer,simultaneously locatingand researching dataand buttons and menus,and then on the fly assemblinga solution to that problem.And this was wasting a lot of timeand a lot of energy for our agents.So after extensive UX researchand many, many prototypes,we've landed on a new UX processand we call it listen match solve.The idea is that theagent will primarily focuson just listening in the customer,and we're gonna surface on the UI screenthe most likely topics for that callso that the agent can just hearwhat the customer is saying,click on what they see on thescreen that matches the topic,and then proceed to diveright into a solution.So we're gonna illustratethis by walking through our UXwith a little bit more detail.So in this UX, we surface themost likely topics for a call.And a topic could be a product order,it could be even a financial product,it could even be a workflow.Either way, we give manyoptions to the agentso that one of them is likely the issue.And once they click on the topic,they can see a summary ofpertinent details about that topicas well as actions thatlet them update this topic.And what we found isit's only so many actionsyou can do for a particular topic,only so many things youcould do for a retail order.And so by narrowing downthe agent position treeto first just find that right topic,it really funnels down theoptions they have to choosebetween really strictlylimiting their cognitive load.But now they found the righttopic, the right issue,how do they solve it?And that's where the step bystep guidance really kicks in,where we give thesesequences of simple UI pagesto the agent to really let them focuson one thing at a time, whetherthere's one input field,one question to ask the customer,they can slowly go step by step,getting the right information they needto help the customers matter.Additionally, along the way,we will feed a bit of agentscripting to the agentso that, let's say theysubmit a transactionand they get a confirmation message,it won't just say thatit's been successful.The success message will read as a scriptthey can read out loudtelling the end customerwhat's just happened and what to expect,like an email notification.Additionally, even whendoing different questionsand inputting fields,with inline standard operatingprocedures and policiesright into the question,so in case a a customerhas a question on the spotabout different options,the agent can read them right out loud,rather having to gofind a knowledge articlein some buried data.Now let's say our topics were wrong.Well, in that case,we always give an escapehatch to the agentto find the right workflow for them.So for example,if the agent clicks onthe find more button here,they'll see a whole list of workflowsthat they have access to.And then now they can figure out,"Okay, which one do I need?"This way we give them thefreedom to always self-serveand find the right solutionin case our predictionis a little bit off.So now I wanna talk about the outcomesof this new UX we've rolled out.First, we've been ableto reduce onboarding timeby down to 50%.And the main reason for thisis the app is so intuitivewe don't have to spendthat much extra timeteaching all the different menu layers.It's pretty much a two differentpanel screen to go through.But more important,we've actually been ableto reduce our time to proficiency by 40%.Agents are using the same UXregardless of the use case they follow,so it really creates asimple streamlined experienceso they can learn as they go.It's the same methodologyof listen, match, solveregardless of the new line of businessthey might be assignedto, or any use case.So they can basically learn as they goand not have to worry aboutdoing retraining over and over.Additionally, we've been able to reduceour average contacthandle time by up to 35%for some departments.And right now we're at a globalaverage of 20% reduction.And perhaps most importantly,we've actually reduced agent heart rates.And I take this really seriously.We did do biometric studieswhere we hooked our agents upto devices and did AB testingof the old UX comparing them to new UX,and the heart rates did go downand self-reported stresslevels also reduced.And while we don't advocate Connectas a healthcare product in this manner,we are really just happy that we're ableto make the daily livesof our agents better.- All right, well this UX processwas research and developedin the context of Amazon customer service.But we can reallygeneralize this UX processand apply it to other types oforganizations and use cases.Here we have more of a genericretail organization example.An agent at a retailorganization is using guidesto help a customer return address.The customer's information,purchase history,as well as payments, are all consolidatedinto a single workflow for the agents.Can we go beyond retail?What about financialservices organizations?Well, in this example,an agent at a credit cardcompany is using guidesto help a customer add anauthorized user to a credit card.New user information entry,approvals and follow ups,again are consolidated into asingle workflow within guides.How about hospitality?In this example, an agentworking for a booking siteis using guides to help acustomer with a new reservation,whether it's one reservation,two reservations,handling payments, it's all consolidatedinto a single workflow within the guide.Now, guides are notjust for voice contacts,guides can be used for chat contactsand the appropriate guidesare surfaced to the agentbased on the context of the chat.But it's not just voicecontacts or chat contacts,you can also use guides with Tasks.Tasks can be used byagents to create, assignand complete tasks inthe same user interfacethat agents accept calls and chats.Again, the appropriate guides are surfacedbased on the context of the taskthat is received by an agent in a queue.Guides can also be used by agentswho are in after-contact work mode.You can see here the contactcontrol panel on the left,the agent is in after-contact work modeand a new tab has popped openso the agent can complete a feedback formas part of their guided experience.Now, the default design that we'vebeen showing you throughoutthis presentationwas really inspired byAmazon customer service.And you see some of thesuccesses and outcomesthat Jack showed you previously.But we know that customerswant customization.And with guides you cancustomize the way they lookinside of our AmazonConnect agent workspace.In this example, you seethat there is a guidein the central paneland we've also persistedall of the available guidesin the right hand panel,in case an agent wants to make a switch.Going even further, guides are flexible.You may want to take guides outof our Amazon Connect agentworkspace and embed themin your custom agentworkspace, or in your CRM.We know that you make your build decisionsand your purchase decisions to fitthe unique needs of your agents.And you can use our APIs tobring a guided experienceinto a custom workspace likeyou see on the screen here.You can see here that we havethe contact control panelon the bottom right,and we have other widgets,like customer informationon the bottom left,queue statistics on the top left,and then customer sentimentscores on the top right.Now whether you use guidesin our Amazon Connect agent workspace,or you use them in acustom agent workspace,the way you build themis exactly the same.- So I had just a question,how can customers build guides?And when we set out tobuild this experience,we could have released a bunch of APIs,but we really wanted to go witha more low code, no code approachthat enabled any persona ina contact center organizationto build these guides.And when thinking abouthow to do this we realizedwe already have a no-code,low-code builder in Connect,and that's Amazon Connect flows.With flows, which launchedwith Connect five years ago,you're able to use this dragand drop GUI to build IVRs,chat bots and routing logic for contacts.But now with the addition ofone new flow block calledthe show view block,you can create UI-basedworkflows to surface to users.And so with this flowblock what happens iswe send instructions toa client application,in our case the agent workspace,determine what to render,and then based on the agent'sactions and what they click,a response is sent back to the flowand then it carries on executing the flow,either doing backend updatesor going to the next view to show.This flow block can be combinedwith any other existing flow blockin order to create branching logicas well as get customer data.So in terms of integrating data,both options are available toboth integrate with Connectas well as get datafrom outside of Connect.So in terms of data thatalready lives in Connect,it's all that IVR datajust ready to be used.This can be surfaced at screen popstrue agent at the start of the call,but also used as conditional logicto figure out where tobranch in the workflows.Additionally, you can use ourout of the box cases flow blockto get case information to an agentand even automate the creation of cases.An agent can go through astep by step guide and ifyour business logic determinesfollow up work is needed,a case can be automaticallycreated on their behalf,tasking around to otheragents without the agenthaving to do the manual or tedious workof all that data entry.And additionally, likewise,customer profile data readilyavailable to the agent,both you and their guides,as well as to give them the abilityto create new customerprofiles for new customers.In terms of integrating datafrom outside of Connect,we use the Lambda flow blockthat already exists in flows today.Traditionally used to do data dips in IVRsas well as in automated tasks,the flow block can now be usedto get data from external systems,whether it's a vendor productor homegrown solution,and then pull that into the agent UIto give them information they need.But also a way to updatethese backend systemsat the end of a transaction.And what this looks like in practice isall this data from disparate sourcesis unified into single UI pages.This way the agent isn't worried aboutwhere customer profiles data lives,or different transaction history.They just get one toolto make the updates.And to them you remove this burdenof figuring out where they need to go.So there's a few otherbenefits of using flowsfor the solution that we wanna call out.The first is ability to transfercontacts between agents.If Agent A is working through a workflowand they get stuck orthey need authorizationfrom their manager to complete it,they can transfer contact to agent B,and agent B in theirscreen will be able to seehow far agent A got andthen carry on with the work.Additionally, you cancreate branching pathswithin your guides that waythere's complex business logicor standard operating procedures.It can all be codifiedright there in the systemthat way the agent doesn'thave to worry aboutreading up knowledge articles and rules.It's right there for themand guiding them throughout the process.Additionally, there's a linkagebetween the workflowsagents do through all flowsand the original inbound contacts.This allows us to updatethe custom attributesassociated with an inbound contact,such as a distribution code,giving you the way toreally sync your data.And on that vein,we also will have audittrails of agent actions.If you go now to a voicecontact or chat contact record,you'll be able to see allthe different workflowsthat an agent performedduring that contact.And when you click on a workflow,you'll get the complete audit historyof every click they took,the steps they performed,and even the datathey're trying to update.So I do wanna reiterate thata guide is really just a flow.It's a knowable contact flow,but it just has this capabilityof sending data to a UIand determining how that UI should look.But there's always benefitsof sharing the databetween the different types of flowsfor transferring betweendifferent contact flows,which will help for both the maintenanceand modularization of your workflows.And additionally, you getthe benefit of being ableto use flow modules to reallycreate discreet units of workand logic that keeps theflow maintenance scalable.But in very particular noteon the serve connection between flows,the way you actually determinewhat that first screenshould be for an agent to seehas to do with settingthe immediate default flowat the server contact.So when you go to yourcontact flows for routing,you'll be able to setone contact attributethat determines which guide to surface.And you can do this based ondifferent branching rules.You could decide, okay,for certain queues,this is the right landing page to show.Or maybe depending on the IVR response,you dive right into a problem.So for example, you'll be ableto either in the situationswhere you have no ideawhat the call's about,you can adopt our listen match solve UXand give the agent a bunch of topics,then figure out, okay,what should I do next?But in the situationswhere you already knowthe exact intent from theIVR, customer history,you'll be able to surface the exact flowand immediately the stepthe agent needs to doright on their screen.This way you get full configurabilityinto how dynamic you wantto make this experienceand to really drive down theagent contact handle time.The last slide we haveabout flows and theconfiguration experiencehas to do with analytics and flow logs.Basically, we've updated our flow logsto now support thetracking of agent actions.So you're able to firstlook at a particular contactand see as to which guide they chose,and then which steps along the wayand given differentdecision trees do they take.But then more so you can viewall the different workflowsexecuted in aggregateas figuring out, okay,what percentage of my contactsdid the agent take thisparticular workflow,and then down the way, how oftendid this branch get chosen?Overall, by given this data,we really hope to allowmanagers, business analysts,and whoever is creating these workflowsto get the insights theyneed to optimize and improvethe workflows by spotting trendsas well as sticking points.Maybe there's a particular stepthat there's a system errorand it always errors out.Or perhaps agents are manually abandoningthe flow of that stepindicating a UI problem.Either way, we want to get the dataright to the manager's hands with somehelpful pre-canned cloud watch queriesthat really allow these personasto update these flows efficientlyand make sure they'remore effective for agents.- Okay, well, customerswho use Connect todayare excited about step by step guides.One such customer here, Unum,helps millions of people every single yeargain affordable access tolife, disability, accident,illness, dental, and visionbenefits via their workplace.Today, they use a custom agent workspacethat their agents have to help customers.They're excited about guidesto consolidate actionsinto a single workflow andtake the number of screenstheir agents use from five down to one.Further, they believe they may be ableto sunset their existing workspaceand save 15-20% on maintenance costswith the Amazon Connect agent workspace.Partners are also excited about guides.Voice Foundry specializes inthe design and delivery ofcustomer experience solutionsbuilt on the Amazon Tech Stack.They're excited about theflexibility of guides.They believe it'll help themdeliver agent experiencesmore rapidly in an evolving marketplace.So that ends the main partof our presentation todayon step by step guides.If what we've been talking about todayhas interested you or excited you,Jack, myself and our colleagueswill take questions around the floorafter we end the presentation.If you are new to Connector are interested inlearning more about Connect,here is a list of other sessionswe have today and tomorrow.I'll leave it up there for a secondif you want to take a noteon a particular session.Otherwise, thank you.Thank you so much for joining us today.If you wanna reach out to Jack or myself,we have our email addresses up here,as well as a QR code you canget our contact information.Otherwise, have a great rest of your day.
