Sophisticated AI-driven sales agents are revolutionising digital sales, offering telcos a scalable, high-conversion solution to meet growing customer demands. The question for telcos is no longer if they should adopt AI, but how. Should they build these agents in-house or partner with specialised vendors
Organisations may believe that keeping the build confined within the company walls is an inherently safer option, allowing for greater control over data, compliance and brand integrity. The reality isn’t so straightforward. The strategy of enclosing an AI sales agent within a ‘walled garden’ might create more risks than it’s intended to mitigate, a false sense of security that often collapses under real-world pressure – leaving telcos vulnerable to compliance failures, subpar performance, and missed opportunities.
The illusion of safety in in-house builds
Building an AI sales agent internally may feel like the best way to safeguard sensitive data and ensure compliance. However, this may not be the case. Telco sales journeys are uniquely complex, requiring strict adherence to regulations on pricing, eligibility, disclosures, and data usage. Without robust guardrails in place from day one, even the most well-intentioned in-house builds can falter.
Compliance testing, content moderation, security auditing, and protection from prompt injections are non-negotiables for any AI deployment. But meeting these minimum requirements is just the start. The real question is: Is the AI agent actually good?
“AI is still a space of experimentation, but significant investment is being made to solve real-world problems. For telcos, the focus should be on proven solutions that deliver tangible results. Sooner or later, CFOs will start asking, ‘What’s the ROI’?” – Daniel Gurrola, Ex Chief Strategy Officer, Verizon Consumer.
To deliver meaningful results, an AI sales agent must go beyond basic functionality. It needs to foster trust, handle objections, and understand consumer psychology. This requires access to rich, actionable data, drawn from real-world sales interactions, to identify customer motivations that extend beyond price and features, such as emotional drivers toward a purchase.
Specialised capabilities like fine-tuned conversation models, behavioural insights, and robust prompting architecture are essential to ensure the sales agent performs consistently and reliably. Without these, the in-house build that seemed safer can quickly become a liability.
Guardrails: striking the right balance
While robust guardrails are essential to ensure compliance and accuracy, overly restrictive guardrails can stifle the sales agents ability to perform effectively. When the AI sales agent is too tightly constrained, false positives risk creating a frustrating experience for the end user, failing to provide helpful answers, misinterpreting intent, or being unable to adapt to nuanced customer needs.
This is where specialised vendors bring a critical advantage. They have the expertise to design guardrails that strike the right balance: tight enough to ensure compliance and prevent errors, but flexible enough to allow the AI sales agent to engage customers meaningfully and drive conversions. By leveraging industry-specific insights and years of experience, vendors can fine-tune their AI sales agent to navigate this delicate balance, ensuring it delivers both safety and performance.
Doubling your chances of success
The challenges of in-house AI development are reflected in the numbers. According to a recent report, internal AI builds succeed only one-third of the time, while projects outsourced to specialised vendors succeed nearly twice as often, with a 67% success rate.
This isn’t an argument for outsourcing everything. Telcos may choose to retain control over critical elements like the front-end user experience, orchestration, and customer data. But partnering with a specialist vendor can significantly reduce risk by providing pre-built compliance guardrails, scalable infrastructure, and access to broader datasets.
These guardrails are critical. They ensure that customer interactions are accurate, compliant, and trustworthy. For example, they prevent the AI sales agent from retrieving unapproved information, using inappropriate language, or sharing outdated pricing. Vendors also offer stress-tested safeguards that evolve alongside industry regulations, helping telcos stay ahead of compliance requirements.
The competitive edge of partnering with experts
One of the biggest advantages of working with a specialist is the scale and diversity of data they bring to the table. External vendors train their models on millions of customer interactions, fine-tuning them to meet customer expectations, guide users through the sales journey, and drive conversions. This level of expertise and refinement is nearly impossible to replicate in-house.
Moreover, partnering with a vendor doesn’t mean sacrificing control. Telcos can still access full transcripts of customer interactions, maintain ownership of customer data, and oversee active management and monitoring. The difference is that they gain the added benefit of proven infrastructure and embedded safeguards, reducing the risk of costly mistakes.
Changing the question from “do we outsource?” to “by how much?”
For telcos navigating the AI frontier, the focus should shift from whether to outsource to determining the right balance. Certain aspects, like customer-facing interfaces, can remain in-house to leverage internal capabilities. But the core of a successful AI sales agent – compliance, scalability, and performance – requires deep expertise and ongoing vigilance.
The safest strategy isn’t building a higher wall around your AI agent. It’s ensuring the garden inside is nurtured, protected, and constantly evolving. By partnering with a specialist, telcos can accelerate time-to-market, reduce risk, and deliver the high-quality customer experiences that drive long-term success.
Pete Sellwood
Pete Sellwood is CTO at 15gifts.
