Why SMBs Shouldn’t Jump the Gun on AI

India has spawned 150+ AI startups that have collectively raised over $1.5 billion since 2020. Small and medium-sized businesses (SMBs) now have a wider range of AI solutions to choose from – each catering to different needs and budgets. So, why is it that many of the SMBs we talk to aren’t seeing the ROI they expected? 

Perhaps it’s time to take a step back and refocus. With so much hype around AI, everyone wants to be an early adopter. But first, it’s important to have a clear vision and roadmap. A well-thought-out approach to AI adoption is more likely to succeed than a hasty implementation driven by FOMO (fear of missing out).

Lessons from the field

Based on my discussions with Indian SMB customers, here’s what I’ve found to be the biggest challenges surrounding AI adoption:

1. Failed POCs (proofs of concepts): With AI, the proof of the pudding is truly in the eating. A mid-market manufacturing customer learned this the hard way. They were very impressed with a fancy demo of a 24/7 AI service agent. But when the implementation actually began, so did the challenges. Objectives weren’t clearly defined. Outputs from the AI were poor. And eventually, the project had to be abandoned.
2. Hallucinations and toxicity: Many AI solutions are essentially wrappers around large language models (LLMs). While they do solve for certain use cases, they’re also prone to hallucinating and producing toxic responses. A marketing agency experienced this first-hand. They were trying to create India-specific visual assets for a campaign; but since the data wasn’t trained on Indian subjects, the output was riddled with inaccuracies. I was surprised to learn that quite a few AI solutions don’t even provide a testing environment to evaluate the output of AI agents.
3. Security risks: Sixty-one percent of customers believe that AI advancements make it more important than ever for companies to protect their data. A small fintech startup knew how critical data security is; but, to avoid lagging behind their peers, they didn’t prioritize security as the very first criteria when evaluating AI solutions.
4. Inadequate post-sales support: A common complaint among SMB CXOs is that AI vendors often lack the scale and bandwidth to deliver consistent service. As these vendors take on more clients, the quality of service tends to slip. Existing clients then have to rely on self-service, or repeatedly chase after the vendor to resolve issues.

Best practises for successful AI adoption

Given the risks and challenges surrounding AI adoption, what can SMBs like you do to maximise AI project success?

1. Establish clear objectives and outcomes: Indian CHROs expect agentic AI adoption to grow 383% by 2027. But before you rush to roll out your own agents, be clear about what exactly you’re trying to solve. Talk to all stakeholders from the ground up to identify gaps, and map out how AI can help fill them. Document everything to guide your AI evaluations.
2. Identify an AI champion: The success of an AI project depends on how well it’s managed. Determine who in your organisation can lead and drive the project. A single point of contact (SPOC) who can rally resources, remove bottlenecks, and keep team members accountable will maximise the project’s impact.
3. Test, test, test: CHROs expect agentic AI to improve employee productivity by 41.7%. But to realise that potential, there has to be a strong AI testing mechanism in place. Regular testing helps ensure that AI’s outputs are accurate and in line with company guidelines. Even if the testing period goes on a bit longer than expected, that’s ok. It’s better than your customers receiving toxic or hallucinatory AI responses that can damage your company’s brand.
4. Prioritise security: Seventy-nine percent of CIOs feel that AI will bring new security challenges. So, while evaluating AI solutions, don’t be afraid to ask vendors how they plan to safeguard your data. Make sure they have sufficient controls in place to prevent unauthorised data access. Also, check that they won’t be using your data to train LLMs. 
5. Ensure a strong support ecosystem: Just like any software solution, AI tools should also come with robust vendor support mechanisms to resolve any issues that occur post deployment. During your AI evaluations, verify that the solution has a well-defined product roadmap and maintenance support.

Agentforce: Why it matters to SMBs

Agentforce isn’t just another layer on top of the Salesforce platform. It’s the next generation of AI that’s fundamentally transforming how small businesses like yours operate. 

Agentforce has everything you need to design, develop, and deploy trusted AI agents that are always-on and able to autonomously handle various tasks. From resolving customer service requests to creating personalised marketing emails, Agentforce helps you get the job done faster and with fewer challenges.

Here’s what Agentforce provides:

1. A low-code/ no-code agent platform: You don’t have to be an AI expert to build autonomous agents on Agentforce. With our easy and approachable agent builder, anyone with a basic knowledge of CRM can create an agent in minutes.
2. A command centre: With the Agentforce command centre, you can monitor and tweak the output and performance of your AI agents to ensure they operate within company guidelines.
3. Einstein Trust Layer: This robust set of guardrails ensures that your data is never used to train LLMs. What’s more, sensitive data is always masked before it’s processed by the AI.
4. Choice of models: Agentforce works with a range of models – from OpenAI, to Google Gemini, Anthropic, and Salesforce AI. You can even connect your own AI model.
5. A robust partner ecosystem: With Agentforce, you get access to our established partner network and Customer Success Group who can help you get the most value from the platform.