Hi Christopher, I just came across your entry and TP and found it an interesting read. I agree with a lot of your points but am not so sure about the angle that AI is an enabler of these CX / BPO firms to move up the value chain. I mean I'm not so sure how true that is and it's hard to tell from actual data right now.
I'm curious what you think of TP today given all the advancements in AI agents since you wrote this. Also, do you think your thesis would apply to the smaller CX firms on that list like TTEC (which btw has a buyout offer from its founder)?
Thanks for your question, Peter. Since writing my article on Teleperformance (TEP), I've refined my thinking on how AI could impact the customer experience (CX) and business process outsourcing (BPO) industry. I see three broad scenarios ranging from worst to best case for TEP. I won’t assign probabilities, but at TEP’s current valuation, even the worst-case scenario doesn’t seem that bad for the investment case.
1. Worst Case: AI Agents Fully Replace Humans in CX Within 10 Years
• Over the next decade, AI agents take over most customer experience roles, leading to a gradual decline in the number of human employees at TEP and a corresponding reduction in revenue.
• TEP's earnings and cash flow would shrink, requiring management to decide whether to return profits to shareholders or reinvest in new business lines.
• Even in a declining business, discounted cash flows still hold value. Assuming a competent management team can reinvest 10 years of earnings at reasonable returns; TEP could still be worth 4-6 times its current earnings (versus current PE of 8.69), mitigating downside risk.
• AI-driven CX solutions would likely still require human oversight as backups for complex or sensitive cases, preserving some level of human employment and revenue for this part of TEP
• TEP has significant data, subject matter expertise, and strong client relationships, this has value and as I said in the article making AI Agents isn’t simply a matter of running an AI model you also require data, subject matter expertise and clients relationships. One example of such a pivot would be it to pivot toward CX technology consulting. Successfully leveraging these assets could help offset losses in the traditional BPO segment.
• Large firms like TEP may have an advantage over smaller CX firms in implementing AI solutions due to their ability to handle compliance, risk management, and data security—key concerns for enterprise clients. As oppose to smaller firms just rolling out a model.
• In addition, TEP provides multiple solutions to many clients some of which are easier to automate than others. Thus it may be easier for clients to stay with TEP as certain parts are automated while others are not, even if their automations isn’t as good as other to begin with.
• A thought-provoking angle is Jevons Paradox: as the cost of customer support drops with AI, usage could increase. Many people avoid calling support due to long wait times, but if they can get instant, effective AI assistance, they may engage more often, potentially leading to expanded business opportunities.
• While this scenario introduces significant uncertainty, TEP’s historical management strength and its current low P/E ratio (~8.69) suggest limited downside.
2. Moderate Case: AI Increases CX Productivity, Replacing Most Humans Over 20+ Years
• In this scenario, AI adoption follows a more gradual trajectory, reducing human-based CX roles over 20 years rather than 10 years.
• AI augments human agents rather than fully replacing them, allowing for more efficient workflows and reducing the need for large-scale layoffs.
• The slower transition provides TEP with more time to pivot its business model, potentially developing high-value AI-driven CX consulting services.
• The longer runway increases the value of discounted future cash flows, making this a more favourable scenario for investors.
• Potential offsetting factors include an increase in overall CX demand, industry-wide outsourcing growth, or market share gains.
3. Best Case: AI is just another Tool, and CX Growth Resumes
• AI enhances CX operations but does not fundamentally replace human roles any faster than past technological shifts.
• TEP’s business continues on its pre-COVID growth trajectory, and fears of AI disruption prove overblown.
• As concerns dissipate, TEP’s valuation multiple could revert to historical levels, leading to significant stock appreciation.
Broader Implications
Beyond TEP, if AI agents become superior to humans in all communication-based roles within a decade, the implications for the broader economy are profound. Sales, customer service, and many other human-centric industries would face widespread disruption, affecting employment on a massive scale. While this isn’t my base case, it’s worth considering from a macro perspective.
Smaller CX Firms Like TTEC
Regarding your question about TTEC, I haven’t really looked at it in much depth but in general. Larger firms like TEP have scale advantages in AI implementation, compliance, and data security, making them better positioned to navigate AI-driven industry changes. However, TTEC’s buyout offer from its founder suggests confidence in its long-term prospects, which could indicate hidden value or a strategic shift.
Overall, while AI presents both risks and opportunities, TEP’s low valuation, historical adaptability, and potential to leverage AI rather than be displaced by it suggest a reasonable investment case with mitigated downside risk.
Thanks Christopher, I just saw your reply. I also had some more time to research and think about the implications of AI on these CX/BPO firms. I think a key assumption with the market consensus (very bearish) on these firms is that they will simply be displaced by agentic AI and the embedded assumption there is the clients of these firms can simply adopt AI tools. I think this is a very questionable assumption as 1) a large amount of IT workloads still remain on-prem, and 2) even if the client is on the cloud, they need consulting help to migrate and maintain their CX systems and I think that's where the CX/BPO players come in / are already coming in.
So I think the revenue and margin pressures at these BPO firms have largely come from one-off and temporary issues like macro uncertainty and delayed decision-making in trying to digest and assess all the new AI capabilities. Meanwhile, it's clear that AI has been an enhancer for BPO agents. Time will tell but if revenue growth start inflecting at theses BPOs this year then that would be a clear sign that AI is not an existential threat in the foreseeable future.
Agreed, moving from using human call centers to AI agents is not a frictionless process and especially when the technology is in its infancy it is going to take time. I would also add the the revenue going down over the past year I would attribute more to contracts made during Covid ending rather than any losses to AI. Which is why I was surprised that they reported increased revenues and profits in their last quarter and the price briefly spiked before going back down and has only gone up again more recently.
The Covid whiplash seems to be very confusing for investors, and not just in the BPO/CX space but across many industries. It's hard to figure out what is just a normal reversion to the mean after a Covid spike and what is structural / secular change.
One phenomenon that I've noticed lately is that the AI chatbots seem woefully bad compared to actual ChatGPT. Even ones that you would expect to be good from the guys with the best tech like Amazon is terrible. I wonder why that is and so far have come to the conclusion that it's a combo of: 1) they put a lot of limiters on the foundational model, 2) they force the response to be very succinct, and 3) they don't allow the bot to make any actual decisions aside from saying No and referring you to more resources.
These limitations seem wholly self-imposed as the actual AI technology as already far surpassed what we're getting in the AI chatbots. So in my mind this begs the question of whether these limitations are a long-term requirement. I'm leaning more towards yes for now because most people just won't read incredibly long responses, and it's the length of the response that often make ChatGPT so good. But if you force it to respond within X number of words, then the response quality declines dramatically. For legal, regulatory and just safety reasons (from customers gaming any AI decision-making system), companies probably will have to continue having limiters on their AI chatbots and forbid them from actual decision-making. So these dynamics should continue to limit the usefulness of AI chatbots.
That's interesting, I should try some chatbots online. My first thought would be the limitations are so that the chatbot doesn't say something its not meant to. I remember reading Air Canada Chatbot was giving customers discounts and the company needed to stand by those discounts they'd been promised from the chatbot
Hi Christopher, I just came across your entry and TP and found it an interesting read. I agree with a lot of your points but am not so sure about the angle that AI is an enabler of these CX / BPO firms to move up the value chain. I mean I'm not so sure how true that is and it's hard to tell from actual data right now.
I'm curious what you think of TP today given all the advancements in AI agents since you wrote this. Also, do you think your thesis would apply to the smaller CX firms on that list like TTEC (which btw has a buyout offer from its founder)?
Thanks for your question, Peter. Since writing my article on Teleperformance (TEP), I've refined my thinking on how AI could impact the customer experience (CX) and business process outsourcing (BPO) industry. I see three broad scenarios ranging from worst to best case for TEP. I won’t assign probabilities, but at TEP’s current valuation, even the worst-case scenario doesn’t seem that bad for the investment case.
1. Worst Case: AI Agents Fully Replace Humans in CX Within 10 Years
• Over the next decade, AI agents take over most customer experience roles, leading to a gradual decline in the number of human employees at TEP and a corresponding reduction in revenue.
• TEP's earnings and cash flow would shrink, requiring management to decide whether to return profits to shareholders or reinvest in new business lines.
• Even in a declining business, discounted cash flows still hold value. Assuming a competent management team can reinvest 10 years of earnings at reasonable returns; TEP could still be worth 4-6 times its current earnings (versus current PE of 8.69), mitigating downside risk.
• AI-driven CX solutions would likely still require human oversight as backups for complex or sensitive cases, preserving some level of human employment and revenue for this part of TEP
• TEP has significant data, subject matter expertise, and strong client relationships, this has value and as I said in the article making AI Agents isn’t simply a matter of running an AI model you also require data, subject matter expertise and clients relationships. One example of such a pivot would be it to pivot toward CX technology consulting. Successfully leveraging these assets could help offset losses in the traditional BPO segment.
• Large firms like TEP may have an advantage over smaller CX firms in implementing AI solutions due to their ability to handle compliance, risk management, and data security—key concerns for enterprise clients. As oppose to smaller firms just rolling out a model.
• In addition, TEP provides multiple solutions to many clients some of which are easier to automate than others. Thus it may be easier for clients to stay with TEP as certain parts are automated while others are not, even if their automations isn’t as good as other to begin with.
• A thought-provoking angle is Jevons Paradox: as the cost of customer support drops with AI, usage could increase. Many people avoid calling support due to long wait times, but if they can get instant, effective AI assistance, they may engage more often, potentially leading to expanded business opportunities.
• While this scenario introduces significant uncertainty, TEP’s historical management strength and its current low P/E ratio (~8.69) suggest limited downside.
2. Moderate Case: AI Increases CX Productivity, Replacing Most Humans Over 20+ Years
• In this scenario, AI adoption follows a more gradual trajectory, reducing human-based CX roles over 20 years rather than 10 years.
• AI augments human agents rather than fully replacing them, allowing for more efficient workflows and reducing the need for large-scale layoffs.
• The slower transition provides TEP with more time to pivot its business model, potentially developing high-value AI-driven CX consulting services.
• The longer runway increases the value of discounted future cash flows, making this a more favourable scenario for investors.
• Potential offsetting factors include an increase in overall CX demand, industry-wide outsourcing growth, or market share gains.
3. Best Case: AI is just another Tool, and CX Growth Resumes
• AI enhances CX operations but does not fundamentally replace human roles any faster than past technological shifts.
• TEP’s business continues on its pre-COVID growth trajectory, and fears of AI disruption prove overblown.
• As concerns dissipate, TEP’s valuation multiple could revert to historical levels, leading to significant stock appreciation.
Broader Implications
Beyond TEP, if AI agents become superior to humans in all communication-based roles within a decade, the implications for the broader economy are profound. Sales, customer service, and many other human-centric industries would face widespread disruption, affecting employment on a massive scale. While this isn’t my base case, it’s worth considering from a macro perspective.
Smaller CX Firms Like TTEC
Regarding your question about TTEC, I haven’t really looked at it in much depth but in general. Larger firms like TEP have scale advantages in AI implementation, compliance, and data security, making them better positioned to navigate AI-driven industry changes. However, TTEC’s buyout offer from its founder suggests confidence in its long-term prospects, which could indicate hidden value or a strategic shift.
Overall, while AI presents both risks and opportunities, TEP’s low valuation, historical adaptability, and potential to leverage AI rather than be displaced by it suggest a reasonable investment case with mitigated downside risk.
Thanks Christopher, I just saw your reply. I also had some more time to research and think about the implications of AI on these CX/BPO firms. I think a key assumption with the market consensus (very bearish) on these firms is that they will simply be displaced by agentic AI and the embedded assumption there is the clients of these firms can simply adopt AI tools. I think this is a very questionable assumption as 1) a large amount of IT workloads still remain on-prem, and 2) even if the client is on the cloud, they need consulting help to migrate and maintain their CX systems and I think that's where the CX/BPO players come in / are already coming in.
So I think the revenue and margin pressures at these BPO firms have largely come from one-off and temporary issues like macro uncertainty and delayed decision-making in trying to digest and assess all the new AI capabilities. Meanwhile, it's clear that AI has been an enhancer for BPO agents. Time will tell but if revenue growth start inflecting at theses BPOs this year then that would be a clear sign that AI is not an existential threat in the foreseeable future.
Agreed, moving from using human call centers to AI agents is not a frictionless process and especially when the technology is in its infancy it is going to take time. I would also add the the revenue going down over the past year I would attribute more to contracts made during Covid ending rather than any losses to AI. Which is why I was surprised that they reported increased revenues and profits in their last quarter and the price briefly spiked before going back down and has only gone up again more recently.
The Covid whiplash seems to be very confusing for investors, and not just in the BPO/CX space but across many industries. It's hard to figure out what is just a normal reversion to the mean after a Covid spike and what is structural / secular change.
One phenomenon that I've noticed lately is that the AI chatbots seem woefully bad compared to actual ChatGPT. Even ones that you would expect to be good from the guys with the best tech like Amazon is terrible. I wonder why that is and so far have come to the conclusion that it's a combo of: 1) they put a lot of limiters on the foundational model, 2) they force the response to be very succinct, and 3) they don't allow the bot to make any actual decisions aside from saying No and referring you to more resources.
These limitations seem wholly self-imposed as the actual AI technology as already far surpassed what we're getting in the AI chatbots. So in my mind this begs the question of whether these limitations are a long-term requirement. I'm leaning more towards yes for now because most people just won't read incredibly long responses, and it's the length of the response that often make ChatGPT so good. But if you force it to respond within X number of words, then the response quality declines dramatically. For legal, regulatory and just safety reasons (from customers gaming any AI decision-making system), companies probably will have to continue having limiters on their AI chatbots and forbid them from actual decision-making. So these dynamics should continue to limit the usefulness of AI chatbots.
That's interesting, I should try some chatbots online. My first thought would be the limitations are so that the chatbot doesn't say something its not meant to. I remember reading Air Canada Chatbot was giving customers discounts and the company needed to stand by those discounts they'd been promised from the chatbot
https://www.cbsnews.com/news/aircanada-chatbot-discount-customer/
But like you said I also imagine they'd need to make it more customer friendly form such as shorter as well.