The AI Project (A Short Story)

Prologue

This is a story after I quit my job in early 2017 to start my own BPM consulting organisation.

I am publishing this in LinkedIn, as this story is about Artificial Intelligence (AI), Business Process Management (BPM), and Operations Excellence (OpEx) in varied levels of detail.

It is a long ‘short’ story – almost 8000 words, so I have structured it in chapters to enable you to pick it up after short (or long) breaks. Hope you enjoy the read.

All names of businesses, persons and locations have been changed to protect privacy and for confidentiality reasons.

.

.

.

The Wait

Just like any other person nurturing an entrepreneurial dream, I had made lots of plans to make it big in my new endeavour. I had gained some new skills while I was still working, added a few more after quitting and started building a good network of like-minded people and organisations. I got into formal partnerships with a few product companies, and with a couple of consulting firms. All very well indeed, but for the fact that it wasn’t paying my bills, at least not in full. My savings had kept me going for some time, but later on I had begun to feel the heat. Not that I was running out of confidence (in fact I was left with quite a lot of that), but I was running out of money. I was getting some odd jobs here and there, but nothing that made me feel I had done the right thing. I was reaching the point where I had to find a long-term client or else turn back to being employed. I tried reaching out to my old friends, ex-bosses, contacts for that start. But that was a time when the contract market was at its low – budget cuts and all.

One afternoon in early Jan 2018, after doing some heavy research in my makeshift office in Bangalore, I was taking a quick snooze when I received a call on my smart phone. Thanks to TrueCaller, I could see the caller name appearing as ‘Snotty HR manager – Don’t Bother’. I don’t remember why but I impulsively answered the call. The caller introduced herself as Lynda, and informed that she had looked up my LinkedIn profile and appreciated a few of my blogs, especially around process technology and automation. She represented a large non-captive, offshore Operations organisation headquartered in New York, and her sponsor was looking for an experienced, techno-savvy professional to lead a small team of operations analysts. Relocation terms and salary was negotiable, but they were keen on a couple of years of commitment. The job description appeared to be hazy, but she suggested that my profile was specifically picked by the hiring manager and that I should at least do the interview. No need to send any updated job profile, she had said. Though I was keener on getting a contract, I was intrigued and also felt good about my profile being hand-picked. There was a good chance that this was a gimmick, possibly to meet her targets, but I thanked the lady and accepted to attend the interview. Coordinates were to follow, but I inquired about the hiring manager details. She said she would email me his profile right away. She had definitely not sounded snotty to me, I had thought and dozed off again.

In the next hour, I had received a Skype call invite (for the next day, 2 pm my time), along with Aidan’s profile link in the company website. The hiring manager’s name was Aidan Murphy, a Director of Special Operations in the company. Apparently quite a senior chap and driving technological innovation in the company’s operations space globally. His office address was in Los Angeles, USA. This looked more interesting than I had realized. I tried looking up his profile on social and professional media but found none listed. Smart fellow, I thought.

The Interview

The next day, I had cleaned up the wall and the lighting in my cluttered office, made sure I had re-started my laptop and freed up my RAM to provide a good call quality. 8 gig RAM and 256 gig SSD were proving quite insufficient nowadays. I had joined the Skype bridge a few minutes ahead of the scheduled time. Aidan joined in at 2 pm sharp!

We introduced ourselves and exchanged pleasantries. Aidan said he was an Irish immigrant in the US, having studied and worked in California most part of his life. He had worked in a wide range of operations and technology roles across industries, and now he was managing a special operations project in this company. He had joined into that call from his hotel room in Södertälje, Sweden. He was travelling there to launch a new operations facility for a healthcare diagnostics company. I tried my hand at a joke about the weather and the daylight there, but he didn’t seem to get that. And the discussion moved on to the main topic of this interview.

I spoke about my experience and achievements, and also about my current company and what I intended to do. Aidan appeared to listen quite intently, though the video was patchy at times, nodding his head and making the right sounds. Aidan stopped me at times to ask relevant details of my experience in operations management, process improvements and theory of constraints. I shared my frank opinion of some of the concepts, how they worked well in some and how they had to be tweaked in others. He nodded his head just once after each of my responses. He asked me a few more specific questions around my people management aspects, for which I answered as honestly as possible. I could not but smile at his somewhat accented English, especially with the ‘Ay’ sounds (being a non-native English speak, I am not very familiar with all English accents, but I had worked with an Irish boss years ago, so I knew that one for sure).

At about the 10th minute into my discourse, he intervened and explained his situation. His company had set up an offshore operations facility in Krakow, Poland three years ago, which was growing very fast. They were looking for a person with a multi-cultural experience in technology and operations to lead a team of about 12-15 operations analysts. This team had been processing applications for opening trade accounts for customers of client financial institutions over the last two and half years. No end-customer contact, liaising with relationship managers of client companies, and entirely image and documents processing. The analysts’ average tenure in the role was about 4 years, with a few fresher management trainees and others ranging between 6 months and 6 years. The analysts were a mix of different nationalities, European, Asian and African but all with prior experience in similar operations.

I was really surprised and asked why an Operations Director was interviewing for this position and that too without the reporting manager, when it was just a first level manager position for what appeared to be typical processing operations. And more importantly whether he had confused my job profile with someone else’s – it hurt my pride having tagged my current position as Founder and Principal of my own company.

Aidan smiled as if anticipating my question and explained. This specific piece of operations though ordinary was a high-priority and confidential project for the company. This team was one among the few identified for a pilot Artificial Intelligence (AI) rollout for the operations. The software was to be implemented on the desktops of these analysts, which would closely observe their work over a year, record and digest all scenarios including the analysis patterns, decisions, errors, rework, etc. The analysts were already using an automated workflow solution that enabled smooth flow of work, decision-prompters and captured timestamps and flow of the end-to-end process. There were some default reports that could be customized with the help of a Dev Ops team.

My role (if I took the job) would be to ensure that the process is streamlined, all major bottlenecks removed, and the team’s efficiency and productivity increased before the AI software is implemented. Setting up the AI training methodology and related operations planning should lead to going live with processing of a new global client’s portfolio of applications. The client was also investing in the AI to drastically increase their customer’s onboarding time and portfolio quality. I would have to oversee the resultant AI operations and gradually show an improved performance over the following year.

This required a manager who was not only good with people management, but one who also understood how the software behind the work actually worked. Aidan said, having gone through a carefully selected list of profiles, my experience in process automation management, technology consulting, operations management and team managerial skills caught his eye and he believed that I fit his requirements quite well.  

His reply appeared to somewhat convince the latter half of my question; I was overly happy though to be appreciated in this manner, but it still did not answer the first half of my question about why he was interviewing me alone and who would be my reporting manager. I was taken aback when Aidan answered somewhat solemnly that I would report directly into him.

I was not unfamiliar working with remote managers, but his role seemed to be very strategic and the org alignment seemed a little awkward. I remembered that there would be other teams selected for this pilot, and possibly their team leads would be reporting to him as well. But how would this work when the other team leads who are not part of the pilot are reporting to a hierarchical structure there? Wouldn’t there be any conflicts? Why not select all pilot teams from the same location and possibly seat them all together so that the team leads could work together? I asked Aidan these questions and he simply shrugged saying that the pilot teams are selected on criteria and they are incidentally spread across different sites. And managing conflicts with other teams and team leaders would be part of my job responsibilities. The goal was to ensure sustained and possibly improved performance of my team members amidst all these scenarios. And, he added, my role reporting to him meant that he expected me to be highly self-dependent and take decisions without involving him. I may not even be able to meet him in person during the pilot as he was mostly travelling, and at times he would be incommunicado as well. Email mostly, messenger and phone sparingly, would be the communication channels.

I was by now quite keen on this role, not just for the technology and international exposure, but also for the different reporting structure it entailed. But since I had set up my own company, I would be able to take up this role if offered as a contract through my company. Aidan had no problems as to how I took up the role, except that the remuneration would be the same as what he would offer a salaried manager. There would be very attractive bonuses every quarter, he had added, based on the project milestones being met. We ended the call agreeing to start on the paperwork and logistics but acknowledging that my company due diligence, visa and travel arrangements would take a few months. I imagined my wife and son being thrilled with the move. We had to prepare to live in Krakow for the next two years! It would be exciting as well as challenging.

The Landing

To cut a long story short, I moved with my family into our temporary new home in Krakow in the middle of April 2018. I had spent the first week in onboarding, HR induction, process induction, introduction with my team members, meetings with my peers in the non-pilot areas and their leadership, and 1-to-1 sessions with my span. I had loved the 16 gig RAM / 1 TB SSD / 8th Gen i7 processor that came with a special issue laptop for this project.

20 analysts were reporting into me instead of the earlier mentioned 15 – five new hires were chosen for the pilot to give a wholesome mix of tenure for the project. My predecessor had moved on the week before, and I went through his hand-over notes to read through later. The team was well aware of the AI project and the change in reporting structure – a project leadership team had walked them through the proposed methodology and had sought their active involvement in order to make it a success. The general communication was that there would be no impact on the analysts’ jobs, and they would be involved in the wider rollout because of their learnings and exposure in the pilot.

I had a couple of video meetings with Aidan in my first week.

The first meeting with Aidan was attended by my span of analysts as well. After introductions and an overview of the project, Aidan enquired about their views on the pilot. There were mixed reactions – some were very keen to be part of a tech-driven initiative, a few were neutral and decided to wait it out, while the more tenured among them wanted to move out as they were skeptical of the proposed org structure. They felt that the absence of a skip-level connect in the site would give their peers in other teams an edge in their performance evaluation, escalations, and visibility to strategy and planning. Aidan assured them that they could email or chat with him directly in case I was not around or for any escalations. He had one specific ask from everyone though in respect of email conversation – to use a specific subject line for any emails to him, as he had built some rules for prioritising his responses while he was travelling. I noted a plus point in my mind for this. He also suggested that they listen to what I had to say before taking any decision, and that I will take good care of them as I was on the ground with them. The ball was already in my court.

I noticed some unusual mannerisms of Aidan while on the video call. His attention was rarely diverted from the camera and even when it did, it seemed deliberate, somewhat like trying to relax his eyes by looking at a distant object. He never interjected till there was sufficient gap in the conversation, and he never responded till somebody mentioned his name before asking him a question. I gave him plus points for these manners. But he used a lot more ‘beg your pardons’ and ‘come again pleases’ than necessary, so much so that I thought he had some cheap headphones on. Minus points, I reckoned.

My second meeting with Aidan was to agree on the AI project rollout plan, the training methodology and performance measurement. As I would be focused on the operations, I suggested a dedicated project manager to drive this. Aidan promised a shared PM resource. We agreed that the key to a successful AI implementation is to first have a relatively stable process. It may not have to be the best process, as the objective of AI is not to make the process perfect but for it to have far less variations than when performed by humans. I was to come up with an operations plan to fit within the overall project. This plan should be re-usable for the other managers in the pilot to implement their part of the projects. The next manager was due to onboard only two months from then. He also cautioned me about the three senior members who wanted out, that they were in the project for their capabilities, and it was my responsibility to turn them around to make the project a success. It was a great strategy, I thought, to have smart nay-sayers in the project, get them to see the works so that they advocate the cause going forward. I was to build the training plan in consultation with the PM.

Towards the end of that meeting, Aidan had disclosed a medical condition that he had. He apparently encountered bouts of ‘temporal lobe seizures’ infrequently. And therefore, he avoided sudden head movements. But whenever he could not help that, such as when he sneezed, the seizure would last anywhere between 30 seconds to 2 minutes, during which he would appear expressionless and unable to speak. He had shared this with me so that I should not be alarmed and just wait for him to come around. I could sing a song for him during that time, he had suggested with a straight face. And in case he had difficulty recollecting what we discussed, he suggested that I quickly summarise what was discussed prior to the seizure. I had acknowledged it, wondering how he managed with people who were not aware of it.

Aidan had connected me with the PM via an email but excused himself from the meeting saying that he was travelling. The PM’s name was Petrus Signy, who set up a meeting for the following week.

Later I emailed Aidan a fairly concrete operations plan for his review. He responded reasonably quickly with a list of risks and dependencies that I had missed out. He texted that he would not be available for meetings for a couple of weeks as he was travelling across Asia Pacific, but that he would respond to emails marked “Top Urgent”.  He had suggested that along with the process I should also focus on the people involved in the process, and that I consider all possible impacts on my team members not being set up like their peer teams.

Aidan had copied his boss on one of the emails as a backup. His boss was Paul Rashid. I had looked up Paul’s profile on social media and the staff intranet. He was a sort of strategy wizard in AI and held a senior director’s position. He did not appear to be active on social media but was highly respected in the tech and academic circles.

The Start

The following two weeks were hectic. I had a series of meetings with Petrus, the PM. He was working on two other smaller projects, but he was available 50% of his time for the AI project. Petrus seemed to be tech-savvy and agreed to do most of the deployment planning himself. He arranged a walk-through of the AI software and its functionality with the development team. The software as well as the project was named Abia – the name apparently had its origin from old Semitic roots meaning ‘to be’ or ‘to become’. I shared the operations plan with them for including as inputs into their overall project.

I also set up a meeting with my team members to discuss the operations plan, and then had separate 1-to-1 conversations. I noted their individual concerns and challenges from a project perspective.   I spent some time understanding and documenting the end-to-end process and noting all possible scenarios including exceptions. I checked with my peers as to what performance measurement system and metrics they used for driving performance. As expected, there was no consistency among the managers and each vertical had their own solution and methodology. I worked with Petrus and the Dev Ops team to customize a reports dashboard for my team’s use. The reports showed historic trends, variation, bottlenecks, comparison of past vs. present, etc. I emailed the draft dashboard for Aidan’s approval, he responded quickly with another set of questions on the rationale and logic of the information. I did more analysis around the questions he asked. Around that time, I used to feel my laptop taking a lot longer to come up with the data I mined, and at other times it was fast as lightning. I was really proud of working on my feature-heavy laptop (the processing speed was top class), but some applications were still taking a long time to open. I did not think much about it at that time, just thought it was a bandwidth issue and ignored the temporary performance lag.

In a follow-up meeting with Aidan – he was back from his trip – I set expectations from him – how he would have to compensate for the missing chain of command for my team on the floor. I made some suggestions – based on my recommendations, he could send appreciation notes to team members based on performance, meeting of project objectives and other specific achievements, and giving them the visibility that they would otherwise be missing. He appeared to note these for his actions.

I also discussed the training plan with Petrus. There were two parts to this training. First, the analysts needed to learn how to use the advanced processing system. Once they were comfortable with it, the AI interpretation layer would be enabled on top of the processing system. While processing a case, the analysts would need to highlight the section they analysed and then select their interpretation of the information analysed. The interpretation layer had pre-decided options such as ‘acceptable’, ‘not acceptable’, ‘inadequate information’, ‘unclear information’, ‘not relevant’, ‘not applicable’, ‘system issue’, ‘policy to be referred’, ‘Expert opinion required’, etc. These would appear on every piece of information they highlighted, and also when flipping between pages, toggling between tabs and switching between applications. In some cases, the analysts would have an option to populate free text to confirm their interpretation. The AI software would then build on existing algorithms to learn what decisions were made from the information and store them for future decision-making.  

As the pilot required the analysts to do the processing while simultaneously capturing responses on the interpretation layer, we agreed to build in sufficient buffer in their performance targets. The buffer would be taken out as the interpretation layers were removed.

After two months of Abia observing the analysts, the process would then be flipped. Abia would start processing the cases and marking responses on the interpretation layer, and the analysts would observe the responses Abia made. The difference was that in addition to observing, they also override the responses in case they were incorrect. This process would be followed for another couple of months till the Abia’s algorithms were further augmented to consider most scenarios and complexities. Any exceptional scenarios or highly complex cases would get excluded from Abia’s processing mid-way if required and handled by the analysts. The interpretation layer would be taken out gradually once the analysts observed consistent reliable outputs from Abia. The quality control (QC) checks would be done by the senior analysts within the team. Any disagreements or misses raised by subsequent business audits would be taken out of the project scope and handled by other teams. There would be a record of such misses.

The Ripples

Aidan was back at his base location by then and began commenting and asking questions based on the dashboard reports. He or his deputy had somehow got access to the dashboard designer and made some changes to the reports. I presumed they knew someone in Dev Ops. The team members who maintained their performance or whose results were even slightly better than before were sent appreciation notes. Those who showed a dip in performance were advised to identify the challenges or bottlenecks and discuss potential solutions with me. I wondered why Aidan was sending the appreciation letters without waiting for my recommendations.

I dropped a note to Aidan asking him to wait for my prompt on the performance analysis. He replied saying that he was inspired by my dashboard and did a bit of digging himself to come up with a few new reports. I was a little annoyed with his initiative, though the new reports were much better than mine, and saved me some time as well. I doubted whether he would be able to do all this by himself – I suspected he had enlisted the support of another project manager to do this work for him. I felt this as an intrusion in my work. I asked for some time to catch with him to discuss this, but he responded saying he was travelling again and would be available only on email and chat.

I asked my team to identify and log the processing issues they faced regularly. I walked them through simple root-cause and pareto analysis tools and taught them to come up with short and medium-term solutions themselves. Most issues they identified were related to IT, information flow and over-governance. I prioritised the urgent issues with the team and raised these as red flags with Aidan in a separate video call. He was travelling in the US and was well past his midnight when he had suggested the call. He suggested that I continue focusing on the people aspects and that he would take care of the governance. I was pleasantly surprised when the process owners cut down on the number of governance review meetings and scheduled only for exception handling. Routine business and good performance were not on the agenda. Aidan must have real clout to get this kind of agreement so quickly, I thought.

My team members were beginning to get nervous by then, for two reasons. Firstly, the AI software would be on their desktops that week. They were afraid that the AI would also automatically highlight their performance issues and bring them negative publicity in a broader forum. Secondly, some of them had inhibitions speaking to me openly about the challenges considering the high visibility and management structure of this team. I advised them to reach out directly to Aidan via email if required. He was their skip-level manager for all practical purposes. The two senior members took the offer and wrote to him asking for separate meetings.  Aidan declined video meetings giving the same reason – that he was travelling but asked them to email him with questions. When they did, Aidan had emailed them back almost ignoring their concerns and encouraged them to use their experience to support this critical project, and that he would definitely ensure they were rewarded. But his email also was quite stern stating that any unwarranted delays to the project will not be viewed nicely. One of the analysts decided to escalate this with local HR.

The Abia project went live within a week of my team members getting trained on the processing tool that were enabled with Abia’s prompts. Cases at different stages of their lifecycle were being observed by Abia. With 20 analysts processing cases at different speeds, day in and day out, there were initial challenges in capturing responses on the interpretation layer. We reviewed the outputs at the end of each day and corrected any incorrect or incomplete responses. After a hectic week of trial and error, the analysts got into the groove quickly and started churning out cases to the satisfaction of Abia tech team.

Going through the dashboard, I identified a few issues with the pipeline and managed to catch up with the source business partner in no time. I found that Aidan was already ahead of me in this aspect as well and had already escalated the matter with the business leadership. He had in fact generated a report from my dashboard and shared it with the partner. I was convinced by now that Aidan had taken on another person to do my job without my knowledge.

The Turmoil

In the second month, I got historical performance data some of my experienced team members’ and compared against their current; I began to see some patterns. Nothing alarming, but there were significant variations. I popped in recurring fortnightly reviews in our calendars, and there began our more interesting conversations. We aligned on the common measurement framework – the key metrics that they would be measured against. Referring to those on a continuous basis, the ask from them was to try and be consistent first with themselves, then with their peer team members and then together uplift the team’s performance as a whole. I set them a steep target of 25% better performance in 3 months, while simultaneously offering to help bridge their gaps. Almost two-thirds of the team members were on my side by now, including one of the three naysayers. But the remaining were still on the fence, not to mention that two seniors were now almost on the verge of rebellion.

Aidan also had set up an urgent project review session with me and his boss. He had emailed me separately saying this call was set up on a priority request for two reasons – one, because of an HR escalation on this critical project, and two, there was a major change in delivery plan.

Getting into a call with Paul was a big plus for someone even at a site lead level, let alone a people manager like me. Paul rarely spoke with anyone other than an operations head; not because he was hierarchical, but mainly because his calendar would be booked for months together. It was my turn to be nervous.

After introductions and courtesies in the call, Paul quickly came to the point. Whether I could deliver the first phase of Abia rollout in 6 months instead of the 12 months as earlier planned. I did not show any visible signs of distress, though my heart had pounded. He had appreciated the fast progress I made in the initial 2 months, and that he had got very good feedback from Aidan. I was not to be too worried about the HR escalation, and that Aidan would take care of it. I voiced my concerns a bit, but agreed that it could be done with all necessary caveats – impact on quality, number of profiles churned, timeliness of decisions, increase in exceptions, etc. Paul got off the call, while Aidan and I fine tuned the plan. We just had four more months to deliver the We shared the plan with Petrus for him to align with the Abia development team.

From month three, Aidan had sent out the performance analysis almost every other day while also raising very detailed questions and asking for urgent answers. I reviewed the data from my dashboard and responded to many of the questions. Sometimes justifying, accepting some of the misses, sometimes denying and at other times challenging Aidan as well. I knew he wanted me to look after the pilot team well as that was a critical success factor for the project, so I had no reason to suspect his intentions. However, I began to look at my work as being reactive to his analysis and not proactive. 

Over the next month, I took more initiative and analysed performance data on a daily basis. I emailed my observations and recommendations to Aidan. He responded with updated analyses of my data, along with counter arguments and measures. I reviewed the reports and found that all his questions were logical – so logical that I suspected that the analysis was being done by an expert data miner. Could this be the same person that Aidan was taking help from, I thought. I was also suspicious whether Aidan had some kind of AI robot deployed on my machine as well. I recalled how my fully loaded laptop appeared to slow down on some days especially when I was working on the dashboard.

I sought time with Aidan to discuss my doubts. I had to be careful lest what I discussed sounded like I was accusing him. He had every right to be careful – he was the project sponsor. In our call, I put forth my case and asked whether he was using any additional data mining or project management support to review my team’s performance. Aidan simply smiled and said that I was reading too much into this. He said he just went through the dashboards like I did and came up with observations. He suggested I continue to focus on the people and ensure a consistent performance.

I had not asked him the AI question then. I did some research to whether any AI can do such analysis and frame logical questions. I came across some examples of how AI can goof up on simple tasks. I had to confirm my suspicions. I had to have some proof that Aidan was using some robotic or AI stuff for my work as well.

It was around that time a funny incident occurred during one of the performance review calls with Aidan. I was discussing a report with him, when there was network error on his side. The video call was still connected, and I could see Aidan clearly, but he could neither see nor hear me. This was possibly the first time we had encountered such a problem. I could hear him asking me whether I was still around. I was not sure whether my response would get through, but I replied nevertheless that I will terminate the call and rejoin. I do not know what he had heard, but I could hear him speak in a very plain voice “The server is temporarily unable to service your request due to system downtime or network problems. Please try again later.” And the screen had gone blank. I thought that was Aidan’s effort at making humour. When I rejoined the call, I joked about what he had said before. Aidan went back to his “beg your pardons” behaving as if he did not remember what he had said. We went back to discussing the reports and had closed the call.

A couple of days later, when I was researching about AI and came across the Turing test, it suddenly struck me that the incident was unreal. My mind went in flashback mode and I recalled all my interactions with Aidan starting with the interview until the latest review call. I noticed how he had successfully created a personality of himself that enabled him to hide his AI weaknesses. His remote environments were all studio-like backgrounds that he could change at will to add credibility, and variety to his work. I remembered how his calls would get disconnected around the twentieth minute or so, and he would join back in a few minutes stating network challenges. He would always ask me to summarise what we had discussed before the interruption. Possibly that was the maximum duration for which he was capable of making intelligent conversations, and that he had to reset himself to continue the conversation. His temporal lobe condition was a clever disguise that he could take advantage of whenever any conversation became complicated. It was a perfect excuse to reboot.

I tried calling Aidan on his phone, but he was not reachable. I sent an email asking to schedule a meeting, I got an out-of-office message saying he was on leave for two weeks. I had struck gold, or dirt, depending on which way I looked at it.

I sent a confidential email to Aidan’s boss, Paul Rashid, giving him a quick narration of my discovery and checked his availability the next day. I had added a postscript that I would be discussing this with the site HR head if he was busy. As expected, a meeting invite was soon in my inbox.

I had waited for about 10 minutes on the call before Paul joined in from New York apologising for the delay. He looked very tired – it must have been a long day for him, as it was late evening for him.  My detailed conversation with Paul had run as follows.

The Confrontation

“I think there is something I should bring up about the Abia project,” I said, staring into the camera without blinking. I wanted to make sure that Paul was not another robot.

“Have you discussed this with Aidan before?” Paul asked.

“Very funny. You know why I asked for this call,” I retorted.

“Do you know that Aidan has gone away on emergency leave?” said Paul as if ignoring what I said. His voice did not seem authentic, it seemed as if he was creating a safety net for himself and the company.

“You know that is not true,” I responded without emotion.

“Why?”

“I think I have been taken for a ride,” I said indignantly. “You of all people would know very well that Aidan was never a real person.”

“You must be imagining things,” Paul exclaimed. “Aidan did mention that you were getting too carried away with the AI work.”

I was beginning to get angry. “Let’s drop the crap and speak like gentlemen,” I said with a quiver in my voice. “I have proof that Aidan as a person never existed. Right from the day he, or should I say it, interviewed me. He is one very intelligent bot, if not anything else.”

Paul appeared to be somewhat uncertain before he continued, “Are you saying that Aidan is a robot? I don’t know what to say. How can you be so sure?”

“You of all people must know about the Turing test,” I retorted.

“I know the test very well. What responses do you have to show that Aidan was not AI, and how do you plan to show them?”

I feigned a smile and replied “I have recorded some of my video calls with him. His ‘beg your pardons’ and attempts at using bandwidth as an issue were a big giveaway. I also have his email responses to my queries on performance analysis. The speed at which he responded with so much of data analysis required superhuman powers. I have never been able to meet Aidan in person, and I am sure that no one else has either. I can see the irony in you choosing his name as well. Aidan starts with AI.”

Paul ignored my last remark. “All those can be explained. Besides, you are lying about recording the video calls. We had disabled call recording features on our communication systems. Regarding you not being able to meet Aidan, it was indeed unfortunate, and I believe he did set expectations around that right in the beginning. Besides, such things cannot be hidden from the HR system. I am aware you are undergoing severe stress on account of the project deliverables, so I can understand if you would like to take a week off and go on a holiday with your family. I will approve the leave in Aidan’s absence.”

I thought about this in silence. I had known that Paul would not be a push-over, especially when he is driving such a critical project. He would have had all his bases covered, and possibly global HR was in on this as well. The offer of a holiday was to make me part of their secret and then start working as if everything was normal. Though it was not a paid holiday, it was nothing but a bribe. If I accepted it, I would end up being morally corrupt just like them.

As if sensing a submission, Paul continued, “Let me play the devil’s advocate in your favour. What if Aidan was indeed an AI leader? Would that not be a technological marvel? No other company has come close to deploying such an advanced level of human interaction till date. You have taken almost six months to be convinced that he was not a human. You accused Aidan of using AI help till about a couple of weeks ago. Now that you have found out the supposed truth, why would you not just enjoy the AI experience and see how to make that work even better?”

I was shocked to hear Paul saying this. “How would you feel if your boss turned out to be a robot now?” I asked him, hoping that would make him angry.

Instead, Paul burst into laughter. “How do you know if he isn’t one?”, he chided. “He acts like one for sure.” I smiled, realizing that Paul was being sarcastic. Paul’s boss was Randy Noble, the Chief Technology Officer of the company. I had heard that Randy usually did whatever the CEO asked without giving too much thought.

“But it is breaking the trust I had built with the company,” I retorted. “I was so passionate about this project, and I believed that I had done so much for it that I expected a lot more honesty and transparency than what has been meted out. I do understand now that this was much bigger than the Abia project. You would not be able to run an AI operations force managed by human managers, so you would have to build AI capability for team management as well.  And there would be a huge cost benefit removing almost four levels of management hierarchy in many of the new operations that would be done by Abia-type technologies.”

“You did not raise the risk of job losses as most people would”, Paul remarked.

“I am very passionate about automation, and especially about AI. I know that it will make redundant many operations that are run the way they are now. But it will also need a workforce to do so many new things that are needed to change the old ones.”

“You are really smart. If Aidan is a bot as you say, this is one more example of what AI can do in hiring the right people. Aidan and I both acknowledge your contribution, and we will definitely make it up to you….”

 “Stop referring to Aidan as if he is a person. At least, after all this,” I cut in.

“I still haven’t said that he is a robot.”

“But you will?”

“Let me continue playing the Devil’s advocate. Do you realise that if we had said that your boss was a robot, you may never have joined our company, let alone working with him / it. Even if we had confided in you after you joined, you would surely have been so biased in your views that you might have made a big hue and cry about this. There was a chance that you might have influenced your analysts to scuttle the Abia project.”

“What can stop me from doing that now?”, I challenged him. “Besides, I am not a problem creator, you knew that when you hired me. I would have supported the leadership AI initiative much better from the inside, rather than as an experimental mouse. It would have the same deal as what my analyst team members signed up for when they joined this project. Aren’t they progressing fine even though they know that their work will be done by robots in future?”

“Listen, I know this call could go on for a long time, and I do not have that much time, at least not today,” said Paul, beginning to get impatient. “Even if we acknowledge that Aidan was AI, let me tell what options you are left with. Option 1, You could go to HR or via the legal route, but please evaluate your position as you have already signed a contract for the project acknowledging that you will be interacting with artificial intelligence, and along with it a non-disclosure agreement as well. You might get into more trouble with our legal team than you can handle,” Paul’s voice had sounded cold.

 “Option 2,” he continued in a mellowed tone. “You could reconcile with the situation and continue on the project as if nothing has changed. You could even accumulate proof and conduct more Turing-type tests to prove that Aidan is not human, which would help us to build more leadership algorithms to prove otherwise. This would continue alongside the Abia project. Option 3, after you have taken some time to think and decide that you cannot do it this way, you could simply resign and walk away. Of course, a gag order will continue to be in place. We will waive any notice clause, and also give you the bonus and other accumulated benefits pro-rata for this year. I can even ask Aidan to give you a good recommendation letter. Let me know which option you would like to choose, and I will arrange the right discussions. Whichever option you chose, you can still take the holiday I suggested. You have earned it.” 

My head was reeling by then. I knew that I could not do much legally as an individual against a giant corporation. I remembered I had not read the fine print of the AI contract, so theoretically I had agreed to be part of the experiment myself. Besides, having gotten exposure at the level of Paul Rashid, there was a much bigger opportunity to do what I wanted to do in AI and BPM. Not to forget that I was getting a week off. A week was like a luxury in these times, I could go on that Scandinavian trip I was planning with my family for some time.

“Can I take a couple of days to think about this and let you know?” I asked, knowing that Paul was dying to get off the call.

“Of course, I am sure you will come up with a wise decision. I look forward to hearing from you soon,” he said and hung up.

I lay still in my chair thinking about the call. I had to discuss the details and options with my family before I took any decision. After all, we had decided as a family to take up this job away from home.

I must have gone to sleep with my headphones still on, when I was jolted by a loud shrieking noise. I looked around trying to clear the cobwebs from my head.

I realised I was still in my makeshift office in Bangalore, and the date on my mobile showed early January 2018.

* * * * *

.

.

.

Epilogue

Since this story now turned out be fiction, the disclaimer is to be read as follows:

“Names, characters, businesses, places, events and incidents are the products of the author’s imagination and used in a fictitious manner. Any resemblance to actual persons, living or dead, or actual events is purely coincidental.”

I am sorry for making you read this far by saying it was my story and ending it on a fictitious note.

When I started to write this, it was intended to be an article on AI implementation methodology linking it with OpEx and BPM methods. But as I began writing and my thoughts transformed to words, I began to visualise what future processes could look like with so much advancement in AI technology, and it turned into this piece.  Nevertheless, I have tried to make the work scenarios look as real as possible.

No matter how complex they were, a lot of scenarios are well set up for automation or AI in future. While there are organisations that currently use AI for basic chat-bot kind of work, there are others who are much deeper into it than just initiating sales, troubleshooting or image processing.

If a robot winning against a world chess champion was a start, and self-driving taxis are almost a reality, then a virtual people manager or a strategy leader is not very distant.  Who knows what algorithms artificial intelligence can write for themselves without the weight of human emotions pulling them down at all times?

Of course, not many companies would want to trick people and sell their technology, like I have depicted myself being fooled in this story. But I am sure there are many examples of organisations taking shortcuts in their pursuit of a larger vision.

At least this story is written by a human. I urge you to try the Turing test via comments in any case. You never know who (or what) could have posted this!

= = = THE END = = =

0 replies

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

Your email address will not be published. Required fields are marked *