Offshoring and AI: How Cost-Cutting Fixation Can Damage Revenue and Brand Reputation

1
312

Share on LinkedIn

Both strategies require careful implementation and constant supervision!

By John Goodman and Peter North, Customer Care Measurement & Consulting

CCMC’s latest National Rage study found that 43% of Americans admit to having yelled at a service person, and over half reported having rage. 9% are seeking “revenge” against a company. Why? The access to and the quality of customer service is the culprit, which is likely exacerbated by carelessly implemented offshoring and AI

Offshoring for back office and customer service functions has existed for decades, with pronouncedly mixed results. Despite showing early promise, the implementation of AI has encountered many of the same pitfalls and stereotypes that offshoring experienced. The following will review the benefits, pitfalls, and success factors for offshoring, and then we’ll draw parallels with the more nascent AI implementations.

Definitions

Offshoring is shifting simply-performed functions (such as back office accounting, account processing, telephone, and email/chat customer service) to lower-cost venues. The bulk of this business has been shifted to low-wage countries where English is widely spoken, including India, the Philippines, Mexico, Peru, and Costa Rica. The wage differential usually provides at least a 50% savings in basic operational costs. Further, given the flexibility of staffing hours around the world, service can be both cheaper and faster.

Artificial Intelligence and Machine Learning broadly consist of response to service and information requests using either rules-based decision guidance or generative composition of documents based on a search of databases, either internal to the company or across the Internet. Savings achieved vs. use of human staff to compose the response can be dramatically higher, as much as 90%. Additionally, decision-making via AI can be lightning-fast, speeding responses to real-time situations. For generative composition, AI can draw upon data and principles from information fields that a human subject matter expert (SME) would never even think to consider.

The Devil is In the Details

If cost and speed were the only relevant criteria, both offshoring and AI would be slam-dunks. However, both have pitfalls leading to errors, reduced productivity, and unsatisfactory outcomes, the causes of which are remarkably similar.

For offshoring

Language

While all the above countries have large segments who speak basic English, accents and idioms present large barriers. For one auto company, I found that at least 90 seconds of every call was wasted on misunderstandings, leading to “I’m sorry, could you repeat that?” in both directions. A second or third repeat of that type of interaction resulted in a total loss of the customers’ confidence and statements of “Am I talking to someone in Mumbai?” or “Can I talk to your supervisor?” I must admit we’ve seen the same type of geographic bias in a U.S. organization (though not as strong a reaction) with call centers which sometimes flowed calls between California and Texas. Satisfaction declined double digits when the CSR was from the other region based on perceptions by the caller that the Texans were “bumpkins”, or the Californians were “flakes”.

Inflexibility and missing nuance and apparently irrelevant details

Staff with a basic understanding of English are often unfamiliar with idioms, the use of sarcasm, or the implied intent of a question or statement. Further, the life experiences of the CSR often lacked key facets of the calling consumer’s environment, e.g., CSRs in the Philippines all took public transportation – none of them owned cars so could not identify with issues of damage, unreliability, or cost to repair. In B2B service environments, a non-professional CSR can have difficulty appreciating the damage to the supply chain manager’s reputation or the operational impact of short shipments or delays in shipping to a hospital or manufacturing facility.

Difficulty recognizing the need to escalate to a more expert domestic system

Escalations to supervisors within a call center happen via the judgment and discretion of the CSR. More localized CSRs who are well-versed in the linguistic, cultural, and other considerations of their recognize “corner case,” “edge case,” or other unusual or unanticipated customer needs, and they can often quickly and confidently escalate the matter. Conversely, offshored CSRs require extensive training and upkeep well beyond the product and process to keep abreast of matters affecting their customers across many time zones, nations, and cultures. Case in point: New Year celebrations, inclement weather seasons, rush hour, and religious/national holidays happen at very different times by country and culture.

For AI, there are surprisingly similar weaknesses

Inability to recognize meaning, intent, and emotion

While AI can certainly make rules-based decisions and inferences from data, it is limited to the learnings it can glean by the quantity, quality, and veracity of the data that informs its every action. It’s a common problem for companies to lack the volume and variety of data to train AI algorithms properly, and these algorithms can also suffer from “data drift” as your current needs evolve and diverge from your past data This can lead to AI gone awry when it comes to interpreting customer decisions. As just one example, AI might conclude that a customer who repurchased a product is loyal and delighted, when in fact they did so because his first purchase didn’t work properly or meet their needs and expectations.

Inflexibility

Response rules are only as flexible as they are designed to be. Recognition of a situation that common sense would make an exception for goes “over the head” of an AI system not programmed to recognize it. They are often constrained to the rules and limitations of their deployment due to legal and compliance concerns, and the finite amount of data from which it can learn. The problem arises because AI is offered as a broad solution when its capabilities are limited. The grey area in between is usually ignored and seldom managed.

Inability to recognize the need to escalate or escape to a more expert human

My previously simple, reliable drugstore interactive voice response system has now been “improved” with a speech recognition AI-driven response system takes me down a rabbit hole 20% of the time. It never volunteers to shift me to a human, just keeps trying to apply the six response paths it has available. A much worse example is the Boeing 737 MAX MCAS system which would not return control to the pilots when standard flight parameters were violated, resulting in two deadly crashes. The takeaway is that AI excels at common and predictable customer needs, but it falls short in the area of unexpected or unanticipated inquiries, and a great many of those arrive via customer support processes.

The Costs of Poorly Applied Efficiency Initiatives.

The challenge finance, operations, and marketing face in evaluating the above noble initiatives is that the cost savings can easily be articulated, but the customer experience (CX) impact often is harder to quantify for two reasons:

Quality problems are not immediately obvious to management and not reported by customers

  • Customers often give up on a call or abandon a chat when dissatisfied and either try again (shop the system) or remain dissatisfied. The iceberg of unarticulated dissatisfaction is often 10-80% for consumers and 30-75% for B2B marketplaces. In B2B, the top reasons for not complaining are a belief that it will do no good or a fear of souring the relationship and provoking retribution.
  • For AI initiatives, most unhappy customers just abandon their efforts with no discernable statement of dissatisfaction. With AI, no news is NOT good news. Even for serious problems, more than half of consumers and businesses do not complain, and for problems that are “just irritating”, between five and 25% of customers complain. Have a banner on your chatbot that says, “We can only solve problems we know about! Tell us if you’re not happy with our response.” IT hates this idea – CX dares to do it.

    A good, but disturbing, rule of thumb is that for each complaint you get about AI, assume there are at least five and maybe 20 other customers with the same bad experience.

Cost savings are immediate while revenue and reputational damage take time to emerge

Reputational damage, including word of mouth and regulatory issues, will emerge over time and only measurement is in place to detect it.
Many companies implement Offshoring and AI initiatives without the requisite quality measurement and supervision, resulting in unnaturally large (false) savings.

  • Across all industries, at least one out of five, twenty percent, of customers encountering a bad experience – will be lost. For a customer worth $1,000 in gross margin, the $20 saved by offshoring is offset five calls that go wrong will be offset by the $200 lost gross margin for one of five customers lost (20% of customers lost X 5 with bad experience X $200 gross margin for $1,000 customer = $200 lost – double the savings of offshoring). Offshore service units seldom have less than 5% of calls with a bad experience so you can do the math. If you handle 20,000 calls per month, that means at least 1,000 customers have a bad experience and you’re losing the revenue of 200 customers per month to achieve your cost savings. Is it worth it??
  • In the above example, the actual ROI is often -200% – that is minus 200%! You save $100 in cost and lose $200 in gross margin. And that does NOT count the negative word of mouth that the experience produces. AI-driven chat systems with a broad front-end invitation, e.g., “How can we help you?”, achieve a maximum of 93% satisfactory response with 7% failures. This can only be rescued and tolerated with a robust failure recognition process that quickly escalates to a human.
  • Companies often implement offshoring and AI-driven chat with a “fire and forget” attitude. Outsourcing actually requires more intensive supervision and measurement than domestic systems because the contracting entity and its supervisory staff have an explicit goal of cost reduction only slightly mitigated by a need for minimal acceptable quality. Damage only emerges well after the transaction and is seldom measured and quantified.

Successful Implementation – Simple But Not Easy

The success factors are strikingly parallel for the two initiatives

For offshoring

  • Only simple predictable transactions – first tier offshore staff are trained to know what they can do and what they can’t. Hyatt Hotels empowers the front line to say, “You need a specialist” and initiate the transfer, which is rapid and seamless.
  • Easy escalation when a negative outcome is identified – AARP has created detailed criteria for recognition of escalation from its main call centers and escalates to VIPDesk, a U.S.-based BPO primarily staffed by home-based middle-aged, unflappable American women.
  • Intense training is needed on recognition of emotion and context, as well as moderation of accents. TELUS, the Canadian phone company, has excelled at the mitigation of accents in its Central American call centers.
  • Intense measurement is required to track both daily operations and specific or specialized KPIs. Measurement tools have obvious importance for any customer support center, regardless of location, but it is even more so when coordinating across continents for continuous upkeep of customer needs and inquiries.

For AI-driven response

  • Only address predictable transactions with effective triage – specify to the customer what path to take. AARP uses AI-driven type ahead to help the customer clearly specify what they desire, thereby assisting triage. Use analytics to identify those issues that will not end well in the hands of AI and connect that customer to a human. 
  • Intense verification and quality measurement
  • Aggressive exiting from any situation that appears questionable. Verint, the consulting firm, has found that when consumers use the phrase, “You people” – damage has been done and the consumer is about to explode – requiring immediate escalation.

Summary

When offshoring and AI-driven response are positioned purely as cost-cutting strategies, disaster is almost guaranteed. This cost-cutting objective is very apparent to customers if they are deployed haphazardly, and it’s rarely appreciated, resulting in real revenue damage and negative word of mouth. 

The solution is limiting workload scope, clear aggressive exiting from the system at the first sign of difficulty, and continuous, intense measurement to surface all the unarticulated system failures.

Unless there is careful planning of the workload to be handled and intense ongoing supervision and measurement, a Boeing 737 Max-like crash is all but guaranteed.

John Goodman is Vice Chairman of Customer Care Measurement & Consulting (CCMC), [email protected]. Peter North is a CCMC Senior Consultant, [email protected].

John Goodman

Mr. Goodman is Vice Chairman of Customer Care Measurement and Consulting (CCMC). The universal adages, “It costs five times as much to win a new customer as to keep an existing one.” and “Twice as many people hear about a bad experience as a good one.” are both based on his research. Harper Collins published his book, “Strategic Customer Service”, in March, 2019. He has also published, “Customer Experience 3.0”, with the American Management Association in July, 2014. He has assisted over 1,000 companies, non-profit and government organizations including 45 of the Fortune 100.

1 COMMENT

  1. Have you come across any business that gives the caller a choice between onshore and offshore call center or between live or AI hone support?

ADD YOUR COMMENT

Please use comments to add value to the discussion. Maximum one link to an educational blog post or article. We will NOT PUBLISH brief comments like "good post," comments that mainly promote links, or comments with links to companies, products, or services.

Please enter your comment!
Please enter your name here