Here is an analysis of 7 functional areas that can leverage AI in the frontline enablement workflow. We get behind the hype to see what works now and what is still work in progress.

This is based on experiments and deployments we have done at Bsharp.

Here is an analysis of 7 functional areas that can leverage AI in the frontline enablement workflow. We get behind the hype to see what works now and what is still work in progress.

This is based on experiments and deployments we have done at Bsharp.

The AI Productivity Mother Lode: The search is ON!

The AI Productivity Mother Lode: The search is ON!

In 2023, the AI ecosystem buzzed with news of groundbreaking innovations. Each of these was promising to transform the world and revolutionize the industry.

From boardrooms to branch offices, everyone was talking about harnessing the power of AI: All to get that additional sale or an additional unit of profitability. 

In 2024, the rubber meets the road. In this note, we focus on Bsharp’s mission: Frontline Enablement. We talk about 7 different functional areas in the line of work of Frontline Enablement and examine how the hype is turning out in reality. We do not look at whole processes but look at the components to understand the bright spots and the chinks. Our Learnings:

  1. AI can greatly aid your workflow and dramatically unlock productivity, just today.

  2. But, no one magic solution can solve all problems, even within a narrow area.

  3. Run experiments. Leverage what works. Train people. Continue experiments.

  4. Measure individual productivity increases and overall organizational impact. Prioritise areas with higher total impact.

  5. Take care of the security and confidentiality of your knowledge bases.

In 2023, the AI ecosystem buzzed with news of groundbreaking innovations. Each of these was promising to transform the world and revolutionize the industry.

From boardrooms to branch offices, everyone was talking about harnessing the power of AI: All to get that additional sale or an additional unit of profitability. 

In 2024, the rubber meets the road. In this note, we focus on Bsharp’s mission: Frontline Enablement. We talk about 7 different functional areas in the line of work of Frontline Enablement and examine how the hype is turning out in reality. We do not look at whole processes but look at the components to understand the bright spots and the chinks. Our Learnings:

  1. AI can greatly aid your workflow and dramatically unlock productivity, just today.

  2. But, no one magic solution can solve all problems, even within a narrow area.

  3. Run experiments. Leverage what works. Train people. Continue experiments.

  4. Measure individual productivity increases and overall organizational impact. Prioritise areas with higher total impact.

  5. Take care of the security and confidentiality of your knowledge bases.

Here are the seven areas:

  • The Promise:

    Paste in the key talking points. Get structured English language content for your training. Democratize training development – anyone who knows the subject can develop content, no perfection in English language is required.

  • Where to leverage:

    Use Generative AI to develop training content from existing material of your product or process. Use GenAI to generate simple slide points and voice over files. Use to develop quiz questions from structured content.

  • Action Ideas:

    Ask your content developer to take an online course on ‘Prompt Engineering’.

  • The gaps:

    Content from AI can get very flowery. Ensure that the final content is absolutely in line with the original content – check for AI’s hallucination. Figure out the security level of the content – the AI might retain and learn from your existing files, factor for it. Also, the quiz questions sometimes do not make any sense – iterate more, add value manually.

  • The Promise:

    Paste in the key talking points. Get structured English language content for your training. Democratize training development – anyone who knows the subject can develop content, no perfection in English language is required.

  • Where to leverage:

    Use Generative AI to develop training content from existing material of your product or process. Use GenAI to generate simple slide points and voice over files. Use to develop quiz questions from structured content.

  • Action Ideas:

    Ask the developer to take an online course on ‘Prompt Engineering’.

  • The gaps:

    Content from AI can get very flowery. Ensure that the final content is absolutely in line with the original content – check for AI’s hallucination. Figure out the security level of the content – the AI might retain and learn from your existing files, factor for it. Also, the quiz questions sometimes do not make any sense – iterate more, add value manually.

  • The Promise:

    Upload a set of files, audio and videos into a knowledge base. The field personnel can ask natural language questions to the knowledge base. They can get instant answers.

  • Where to leverage:

    Use it to answer laborious questions like, what are the ports in this model of PC you offer? The user can use this as a ready reckoner.

  • Action Ideas:

    Use specifically curated Knowledge Base for this purpose – do not use generic knowledge base as they might have table formats (with merged cells etc.) that the AI will not easily understand. Mixing up different types of documents in the same knowledge base might give misleading answers. Get feedback about the quality of the answer – so that it can be improved. Also, understand where the AI is not able to provide answers.

  • The gaps:

    The response times from most LLMs is still high at about ~10-15 seconds. The field team needs to get used to the response times. The answer is mostly in English – local language translations are not reliable.

  • The Promise:

    Upload a set of files, audio and videos into a knowledge base. The field personnel can ask natural language questions to the knowledge base. They can get instant answers.

  • Where to leverage:

    Use it to answer laborious questions like, what are the ports in this model of PC you offer? The user can use this as a ready reckoner.

  • Action Ideas:

    Use specifically curated Knowledge Base for this purpose – do not use generic knowledge base as they might have table formats (with merged cells etc.) that the AI will not easily understand. Mixing up different types of documents in the same knowledge base might give misleading answers. Get feedback about the quality of the answer – so that it can be improved. Also, understand where the AI is not able to provide answers.

  • The gaps:

    The response times from most LLMs is still high at about ~10-15 seconds. The field team needs to get used to the response times. The answer is mostly in English – local language translations are not reliable.

  • The Promise:

    Upload your training content, and your voice over script in English. Get it translated into any language of your choice. Use it readily.

  • Where to leverage:

    To get the first draft of the translation, in an Indian language. To use it in popular world wide languages like Spanish, French, Portuguese.

  • Action Ideas:

    Collaborate with an Indian language expert to validate and simplify the final translation.

  • The gaps:

    Most AI translations of Indian languages are very formal – old textbook language. It will not go well with the millennial audience that might consume your training. Hence, have a human moderator who can use the content as the first draft.

  • The Promise:

    Upload your training content, and your voice over script in English. Get it translated into any language of your choice. Use it readily.

  • Where to leverage:

    To get the first draft of the translation, in an Indian language. To use it in popular world wide languages like Spanish, French, Portuguese.

  • Action Ideas:

    Collaborate with an Indian language expert to validate and simplify the final translation.

  • The gaps:

    Most AI translations of Indian languages are very formal – old textbook language. It will not go well with the millennial audience that might consume your training. Hence, have a human moderator who can use the content as the first draft.

  • The Promise:

    Any text can be converted into attractive voice overs – to be used in the training voice tracts.

  • Where to leverage:

    For most languages you now have a host of AI voices that can translate text to voice.

  • Action Ideas:

    Your content developer can take a course in ‘Simple Speech MarkUp Language’ to develop voice with intonation. They have to develop a method to get complex pronunciations and abbreviations right.

  • The gaps:

    Most of the non-English voice overs still have a robotic, steely tone. This can work for many of your weekly training sessions. Want a perfect voice for a super important training? Use human voices, rather than AI voice overs.

  • The Promise:

    Any text can be converted into attractive voice overs – to be used in the training voice tracts.

  • Where to leverage:

    For most languages you now have a host of AI voices that can translate text to voice.

  • Action Ideas:

    Your content developer can take a course in ‘Simple Speech MarkUp Language’ to develop voice with intonation. They have to develop a method to get complex pronunciations and abbreviations right.

  • The gaps:

    Most of the non-English voice overs still have a robotic, steely tone. This can work for many of your weekly training sessions. Want a perfect voice for a super important training? Use human voices, rather than AI voice overs.

  • The Promise:

    Most videos are made with English voice overs. In a country like India, only 17% of people understand spoken English. For the rest, the video is not of any impact. With AI you can add subtitles in multiple languages. It will increase the reach of your videos.

  • Where to leverage:

    AI can now accurately add the native language (English in case of English videos) sub-titles, in accurate timings.

  • Action Ideas:

    Add voice over in English. Edit as required to reflect the correct words in the video.

  • The gaps:

    The quality of translation of sub-titles is still very formal, and does not accommodate the context and tone of the video. Translate the video in other languages. Get a human reviewer who can edit the sub-titles as required, in the case of mission critical training.

  • The Promise:

    Most videos are made with English voice overs. In a country like India, only 17% of people understand spoken English. For the rest, the video is not of any impact. With AI you can add subtitles in multiple languages. It will increase the reach of your videos.

  • Where to leverage:

    AI can now accurately add the native language (English in case of English videos) sub-titles, in accurate timings.

  • Action Ideas:

    Add voice over in English. Edit as required to reflect the correct words in the video.

  • The gaps:

    The quality of translation of sub-titles is still very formal, and does not accommodate the context and tone of the video. Translate the video in other languages. Get a human reviewer who can edit the sub-titles as required, in the case of mission critical training.

  • The Promise:

    Call centre auditors listen to call transcripts manually, and rate them against specific quality parameters. AI can now listen to the voice calls, rate them and provide processed inputs for auditors to value add. Hence, more calls can be analyzed with the same number of auditors.

  • Where to leverage:

    AI can help you to transcribe the call in both the directions (Speaker 1, 2 etc.). AI can also rate the tone and the sentiment of the call. It can answer specific (simple) questions about the conversation.

  • Action Ideas:

    Run a pilot in your call centre.

  • The gaps:

    When it comes to multi-lingual conversations (Hinglish, or Tamilish), the AI transcription and the ratings are not accurate. It requires human intervention. Also, call centres have specific processes for specific scenarios. This has to be documented in the knowledge base. The AI is not good in understanding the exceptions.

  • The Promise:

    Call centre auditors listen to call transcripts manually, and rate them against specific quality parameters. AI can now listen to the voice calls, rate them and provide processed inputs for auditors to value add. Hence, more calls can be analyzed with the same number of auditors.

  • Where to leverage:

    AI can help you to transcribe the call in both the directions (Speaker 1, 2 etc.). AI can also rate the tone and the sentiment of the call. It can answer specific (simple) questions about the conversation.

  • Action Ideas:

    Run a pilot in your call centre.

  • The gaps:

    When it comes to multi-lingual conversations (Hinglish, or Tamilish), the AI transcription and the ratings are not accurate. It requires human intervention. Also, call centres have specific processes for specific scenarios. This has to be documented in the knowledge base. The AI is not good in understanding the exceptions.

  • The Promise:

    Can brands extend their reach in the retail outlets with a virtual salesperson who can answer the first questions about the product, before handing it over to the sales person – online or in store.

  • Where to leverage:

    AI can help you 1. Ask a set of questions to customers and get their answers 2. Recommend products 3. Answer further questions 4. Handover the process to a human (online or instore).

  • Action Ideas:

    Run a pilot in your most controlled retail outlet.

  • The gaps:

    Things work well as long as the conversation is in English – doesn’t work effectively for multilingual conversations. The response time of LLMs is also low, and sometimes prevents live conversations. Need to manage the knowledge base in such a way it caters to a range of customer situations.

  • The Promise:

    Can brands extend their reach in the retail outlets with a virtual salesperson who can answer the first questions about the product, before handing it over to the sales person – online or in store.

  • Where to leverage:

    AI can help you 1. Ask a set of questions to customers and get their answers 2. Recommend products 3. Answer further questions 4. Handover the process to a human (online or instore).

  • Action Ideas:

    Run a pilot in your most controlled retail outlet.

  • The gaps:

    Things work well as long as the conversation is in English – doesn’t work effectively for multilingual conversations. The response time of LLMs is also low, and sometimes prevents live conversations. Need to manage the knowledge base in such a way it caters to a range of customer situations.

At Bsharp, we are constantly running experiments with our customers in the above hot areas. These will help you accelerate processes, unlock productivity and extend reach.

At Bsharp, we are constantly running experiments with our customers in the above hot areas. These will help you accelerate processes, unlock productivity and extend reach.

Want us to do a workshop with live examples? Want to run a pilot in your organisation.

Please message me at gopal@bsharpcorp.com or reach out through the link below.